Sketch-Based Facial Modeling and Animation: An approach based on mobile devices

FACULDADE DE ENGENHARIA DA UNIVERSIDADE DO PORTO Sketch-based Facial Modeling and Animation: an approach based on mobile devices Ana Luísa de Vila Fernandes Orvalho Mestrado Integrado em Engenharia Eletrotécnica e de Computadores Supervisor: Prof. Augusto Sousa (FEUP) Co-supervisor: Prof. Verónica Orvalho (FCUP) October 4, 2013 c Ana Orvalho, 2013 Resumo Transmitir emoções credíveis à audiência através da animação facial de personagens virtuais requer um nível elevado de realismo. No entanto, este grau de qualidade é muito difícil de atingir dado que o software tradicional para modelação e animação, apesar de muito poderoso, é também extremamente complexo, já que oferece ao utilizador uma grande variedade de elementos de interface geralmente referidos como WIMP (Window,Icon,Menu,Pointer). Dada a sua longa curva de aprendizagem, o uso deste tipo de software está apenas ao alcance de artistas experientes. Mas, com a evolução das interfaces de utilizador, novas abordagens ao modelo de interação humanomáquina provaram ser alternativas válidas ao paradigma WIMP. Estas alternativas, conhecidas como interfaces naturais de utilizador, procuram possibilitar que os utilizadores interajam com os computadores da mesma forma que interagem com o mundo que os rodeia, tirando partido de capacidades como o toque, a fala, a escrita manual e o movimento ocular. Esta dissertação propõe o uso de uma abordagem com recurso a desenhos para manipular a estrutura de controlo de modelos 3D, conhecida como rig, através de um tablet com capacidade multitoque, que não requer qualquer dispositivo de entrada de dados para além dos dedos do utilizador. Esta abordagem, baseada num algoritmo já existente que recorre a simples desenhos, pretende explorar as novas possibilidades oferecidas pelos dispositivos móveis que permitam simplificar significativamente o processo de animação, tornando-o acessível a utilizadores sem experiência nesta área. A solução proposta foi especialmente orientada para o iPad, da Apple, e desenvolvida usando o motor de jogo Unity3D para que se integrasse de forma simples no protótipo do projeto LIFEisGAME que inspirou este trabalho - um jogo sério que pretende ajudar crianças com Desordens do Espetro Autista (DEA) a reconhecer e expressar emoções através de expressões faciais, de forma interativa e creativa. i ii Abstract Realistic facial animation is essential so that virtual characters can convey believable emotions to the audience. However, it is very hard to achieve since traditional modeling and animation software, although very powerful, is also extremely complex, providing to the user a series of interface elements commonly referred as WIMP (Window, Icon, Menu, Pointer). As they require significant learning time, these types of software are only accessible to experienced artists. But with the evolution of user interfaces, new approaches to the human-machine interaction model have proven to be valid alternatives to the WIMP paradigm. These alternatives, know as natural user interfaces, aim to enable users to interact with computers in the same way as they interact with the world, taking advantage of abilities such as touch, speech, handwriting and eye-gazing. This dissertation proposes the use of an hybrid sketching and dragging approach to manipulate a model’s control structure, know as rig, through the use of a multi-touch capable tablet that requires no intermediate input device other that the users’ fingers. This approach, based on a pre-existing sketching algorithm aims to take advantage and explore new possibilities offered by mobile devices to significantly ease the animation process, making it accessible to non-expert users. The proposed solution was specially oriented for Apple’s iPad and developed using Unity3D engine so it seamlessly fits in the LIFEisGAME project prototype, that inspired this work - a serious game that intends to help children with Autism Spectrum Disorder (ASD) to recognize and express emotions through facial expressions in an interactive and creative way. iii iv Agradecimentos Começo esta lista de agradecimentos, como não poderia deixar de ser, por dirigir uma palavra de apreço ao meu orientador, Professor Doutor Augusto Sousa, pela sua disponibilidade e todo o acompanhamento prestado ao longo do trabalho mas, especialmente pelo incentivo que me deu no momento certo, e à minha co-orientadora, Professora Verónica Orvalho, por toda a ajuda e, em particular, por me motivar e conseguir ver sempre um lado positivo nas situações. Por todo o apoio que recebi, o meu muito obrigada! De seguida, gostaria de agradecer à equipa do PIC (Porto Interactive Center), que me recebeu tão bem, e a todas as pessoas que participaram nos testes de validação da aplicação desenvolvida, pelo tempo que disponibilizaram e por todos os comentários que deixaram nos questionários e que, mais que uma ajuda, foram um incentivo para continuar. Já num campo mais pessoal, gostaria de agradecer aos meus colegas de curso, mas em especial àqueles que se tornaram grandes amigos, por ouvirem os meus lamentos e por estarem sempre ao meu lado, com uma palavra de incentivo ou, pelo menos, com uma pequena distração que me fizesse depois voltar com mais ânimo ao trabalho. Claramente que os agradecimentos mais fortes vão para a minha família. Agradeço aos meus tios e primas Toni, Irene, Sofia e Raquel, por serem os meus segundos pais e irmãs mas obviamente, agradeço em especial aos meus pais, que me abriram o caminho que agora termino e que estiveram sempre ao meu lado para me ajudar a enfrentar todos os obstáculos. Obrigada por todo o amor e carinho que sempre me deram mas, sobretudo, obrigada por estarem sempre aqui nos piores momentos (e houve bastantes!). Agradeço também ao meu irmão, Bé, por sempre ter sido o meu maior exemplo. No mesmo patamar da família não poderia deixar de agracer ao meu namorado, Tiago, por ser o meu companheiro em todas as situações e por ter aguentado firmemente mesmo quando lhe "fiz a vida negra". Obrigada pelo teu amor e pela tua infinita paciência. Ana Orvalho v vi “Animation can explain whatever the mind of man can conceive. This facility makes it the most versatile and explicit means of communication yet devised for quick mass appreciation. ” Walt Disney vii xiv LIST OF TABLES Symbols and Abbreviations 2D 2 Dimensions 3D 3 Dimensions API Application Programming Interface ASD Autism Spectrum Disorder AU Action Unit AVO Audiovisual Object CAD Computer-Aided Design CG Computer Graphics CLI Command-Line Interface CoLab International Collaboratory for Emerging Technologies DEA Desordens do Espetro Autista FACS Facial Action Coding System FAP Facial Animation Parameters FAPU Facial Animation Parameter Unit FCT Fundação para a Ciência e a Tecnologia FFD Free-Form Deformer FP Feature Point GlovePIE Glove Programmable Input Emulator GUI Graphical User Interface HCI Human-Computer Interaction HD High-Definition IEC International Electrotechnical Commission ISO International Organization for Standardization LED Light-Emitting Diode LCD Liquid Crystal Display LIFEisGAME LearnIng of Facial Emotions usIng Serious GAMEs LTE Long Term Evolution MERL Mitsubishi Electric Research Laboratories MPEG Moving Picture Experts Group MoCap Motion Capture NUI Natural User Interface NURBS Non Uniform Rational B-Spline OEM Original Equipment Manufacturer PC Personal Computer ppi Pixels per inch RFFD Rational Free-Form Deformer RGB Red–Green–Blue RPC Remote Procedure Call xv xvi SYMBOLS AND ABBREVIATIONS SBIM Sketch Based Interfaces for Modeling SDK Software Development Kit SketchFACE Sketch-based Facial Animation Collaborative Environment TV Television UbiComp Ubiquitous Computing UI User Interface USB Universal Serial Bus UT University of Texas WIMP Window, Icon, Menu, Pointer XML Extensible Markup Language Chapter 1 Introduction Freehand sketching has always been a crucial part of everyday human interaction. It has been used throughout time to communicate ideas to others and to help the thinking process, clarifying one’s thoughts. As it is a natural and intuitive means of communication, sketching is a promising tool to ease the human-computer interaction paradigm. This is particularly interesting for the field of facial animation, once it can help non-expert users to easily model and animate characters. Simple free-hand strokes can replace traditional buttons, sliders and keyboard shortcuts as an alternative of input to control modeling and animation applications. This document presents an approach for facial deformation and animation based on a hybrid sketching and dragging control system that takes advantage of the intuitive interface provided by mobile devices. The system was developed for the iPad and implemented using Unity3D engine. By simply drawing strokes on a control canvas on the iPad, a non-expert user can create facial expressions on a 3D model on-the-fly and then animate the created pose sequence. The results can be visualized both on a small section of the iPad screen or, if desired, in a much larger external display connected to the mobile device. The possibility of integrating other visualization displays helps the interaction between multiple users that is also supported by the system, creating a simple and intuitive collaborative environment for modeling and animation. This chapter provides relevant background information on this dissertation. It presents the main challenges addressed, the motivation for the developed work and the objectives and contributions to achieve and briefly describes the proposed solution. 1.1 Motivation Interactive devices with pen-based interfaces have become popular in recent years and tablets and touch screens are now common input technologies for many applications [8]. Sketching, as an intuitive means of communication, provides a natural alternative to the multi-tool selection paradigm of traditional interfaces, being now used in a variety of domains such as front ends for Computer-Aided Design (CAD) systems, automatic correction or understanding of diagrams for 1 2Introduction immediate feedback in educational settings, alternative inputs for small keyboard-less devices, or gestural interfaces. In the context of facial animation, a sketch-based interface can significantly ease the manipulation of the virtual characters, opening new opportunities for non-expert users that usually take a lot of time to master the powerful but complex interfaces of traditional modeling and animation applications. This dissertation is inspired by the LIFEisGAME project that intends to help children with Autism Spectrum Disorder (ASD) to recognize facial emotions in an interactive and creative way. ASDs are a group of development disabilities whose symptoms include problems with social interaction, communication (verbal and/or non-verbal), restricted activities and interests, both in children and adults. The severity of the symptoms, their combinations and patterns of behaviour significantly varies from patient to patient. 1.1.1 The LIFEisGAME Project LIFEisGAME (LearnIng of Facial Emotions usIng Serious GAMEs) [9] is a project funded by Fundação para a Ciência e a Tecnologia (FCT) [10] under the program UT Austin|Portugal, also known as the International Collaboratory for Emerging Technologies (CoLab) [11], in partnership with Faculdade de Psicologia e Ciências da Educação da Universidade do Porto, University of Texas in Austin, Instituto de Telecomunicações [12] and Microsoft. It explores a new and innovative approach that uses a serious game to teach children with ASDs to recognize and express facial emotions. The main objective of the LIFEisGAME project is to deploy a low cost real time facial animation system embedded in an experimental game that enables further study of the facial emotion recognition challenges of these children and also allows to analyse if the use of virtual characters in interactive training programs can help the rehabilitation of ASDs patients. Figure 1.1: LIFEisGAME attempts to show how it is possible to apply a serious game approach to teach people with Autism Spectrum Disorder (ASD) to recognize facial emotions. 1.2 SketchFACE Overview 3 The official presentation of the project took place on June 11th, 2013 where a prototype application was shown and explained, including four game modes that correspond to different learning cycles: Recognize Mee, encourages children to watch a sequence of random facial expressions and recognize the emotion; Sketch Mee, where the user constructs a facial expression on a 3D character to match a defined emotion; Mimic Mee, where first an avatar mimics the players facial expressions and then the user must follow the avatar’s expressions and Play Mee, where the player is required to perform expressions that correspond to a situation depicted in a story. The work of this dissertation focused on a specific module of the LIFEisGAME project that uses sketch-recognition techniques to allow the player to draw facial expressions on an avatar, modifying its appearance. The core of this module consists on a facial sketching control system developed by Miranda et al. [13] that allows complex 3D deformations with just a freehand drawing on a 3D mesh or on a virtual canvas. Based on this, a new application was created - SkethFACE (Sketch-based Facial Animation Collaborative Environment) - to allow collaborative work on the modeling and animation processes. SketchFACE explores networking concepts to allow multiple users to simultaneously deform a model in a joint effort to produce an animated sequence of poses and supports multiples devices for better content manipulation and display, increasing the immersiveness of the animation environment. SketchFACE aims to extend the scope of the LIFEisGAME project to individual without ASDs by providing a simple and intuitive collaborative facial animation system appropriate to non-expert artists and people with no modeling background that can range from young children to multidisciplinary teams where those in charge need a simple tool to transmit their ideas to the ones who will execute them. 1.2 SketchFACE Overview SketchFACE (Sketch-based Facial Animation Collaborative Environment) is an easy application for modeling and animation of 3D characters controlled with simple strokes on 2D canvases that represent each region of a model face. It provides a collaborative environment where multiple users can, simultaneously, deform a character, in a joint effort to create facial expressions that can be saved to a timeline and animated. It was designed to run on mobile devices with a prototype deployed for Apple’s iPad. The tablet works as the control interface to manipulate model’s geometry but the results of the deformations can be seen in a much larger HD TV, if one is available and connected to the iPad via Apple TV. This way, SketchFACE supports multiple users and multiple visualization displays, resulting a simple but immersive tool to help facial modeling and animation. 4Introduction Figure 1.2: SketchFACE Pipeline. Main Features The application allows the user to: •Create different expressions on a 3D character simply by dragging control points or drawing strokes on 2D canvases like sketching on a paper; •Generate an animated clip with the expressions created and saved; •Connect an external display to the iPad so that it becomes the control for the content shown in a much wider screen (like a TV or a computer screen). Share with other simultaneous users or with an audience the poses created or the resulting animation; •Use it alone or team up with some friends to make the animation process more fun and interactive. 1.3 Objectives This work aimed for the accomplishment of the following objectives: O1. To carry out research on facial animation techniques with special focus on sketch-based animation interfaces that leads to a timeline-based facial animation application based on the facial 1.4 Document Outline 5 sketching control method developed by Miranda et al. [13]; O2. To carry out research on interactive devices technology and the new trends in user interfaces so that the final application suits the chosen implementation platform: the iPad; O3. To explore different models to define the most adequate user interface to allow an easy and intuitive manipulation and animation of 3D characters. O4. To deploy a functional prototype application of the system, for the iPad, that includes a standalone version, for single users, and a collaborative environment, allowing simultaneous alteration of a single model. O5. To validate the created application with a set of usability experiments. 1.4 Document Outline The remaining chapters of this dissertation are organized as follows: Chapter 2: State of the Art - Presents some background content divided into two major sections: Character Facial Animation and Sketching Interaction. The first one discusses the complexity of facial animation, briefly describes the stages of a traditional animation pipeline and the major techniques used in each one. The second one presents an evolution of user interfaces and a study of emergent technologies related to interactive devices. It also details a new approach in the human-computer interaction model - sketch-based interfaces - and how it can be used to ease the modeling and animation processes. Chapter 3: SketchFACE: Overview of the Proposed Solution - Defines the proposed solution and details all the preliminary work that led to the development of SketchFACE with special focus on the problem statement and the main system design decisions. Chapter 4: SketchFACE: Implementation - Describes the implementation of the proposed modeling and animation system based on an hybrid method of dragging and free-hand sketching that allows collaborative work to produce facial expressions. Chapter 4: Results and Validation - Describes the experiments conducted with users to validate the application and to reach some conclusions about the value of the mobile device based approach and of the collaborative environment. 6Introduction Chapter 5: Conclusion and Future Work - Summarizes the work developed in this dissertation and presents suggestions for future work related to this topic. Chapter 2 State of the Art This chapter, that provides some relevant background content, is divided into two major sections. The first one, Character Facial Animation, begins by exposing the challenges raised by realistic facial animation. Then, it describes the traditional animation pipeline and the major techniques used in some stages and the section ends with a brief overview of the major standards used to categorize facial expressions. The Sketching Interaction section presents the three phases of interface evolution, focusing on natural user interfaces and addressing some emergent technologies related to interactive devices. The section ends with the description of the main stages of the sketch-based interfaces pipeline, showing how this new human-computer interaction model can be used to ease the modeling and animation processes. 2.1 Character Facial Animation Facial expressions are essential to convey emotions to a character. But creating realistic facial movements is complex due to the intricacy of the human facial anatomy, with all its subtle textures and small wrinkles, and to people’s inherent sensitivity to facial appearance. Convincing animations demand quality models, manipulated by experienced artists through appropriate control interfaces and extensive and time-consuming manual tuning. 2.1.1 Facial Animation Facial animation began in the 1970s with the pioneer ideas of Frederick Parke that produced the first computer generated animated sequence of a human face changing expressions [14]. It remains one of the most fundamental problems in computer graphics, often separated from regular character animation due to its higher complexity. The difficulty to achieve believable animations arises mainly due to the morphology and behaviour expected from a facial model. According to Richie et al. [15], facial style can be defined as hyper-realistic, if it looks realistic and fantastic at the same time, photorealistic, designated by the authors as no-style, and higly-stylized, if it can range in size and shape regardless of the constraints of the physical world. The performance of each virtual character must be in tune with its appearance since it is key to achieve believability: 7 14 State of the Art Figure 2.7: Three examples of viewport-based UI. a) 2D controls by Alexander et al. [4]; b) 3D controls by Komorowski et al. [5]; c) 3D controls by Grubb [6] 2.1.1.3 Animation Techniques Once a model’s geometry is defined and the control structure that allows its deformation is created, the character can be animated. Many different techniques have been developed for this process but, in general three major approaches can be identified: keyframe interpolation,motion capture and procedural animation. Since they are rarely used individually, these approaches can be complemented by physically-based and geometrically-based techniques. Keyframe Animation Keyframe animation is the easiest and oldest completely geometric technique that offers an intuitive approach to facial animation: several complete face models with the same topology, called keyframes, are created for a given set of points in time and the in-between frames are obtained by interpolation of these keyframes. Realistic animation requires a good number of keyframes. If not enough are used, the in-betweens will be too unpredictable, the path of action will usually be incorrect and objects may intersect one another, demanding for exhaustive reworking of intermediate poses [35]. The simplest case of keyframe animation to be mentioned corresponds to an interpolation between two keyframes at different positions in time (figure 2.8). Figure 2.8: Two different poses and the resulting interpolation. Left: Neutral pose, Right: ”A” mouth shape, Middle: Interpolated shape. [1] 2.1 Character Facial Animation 15 Due to its simplicity, linear interpolation is commonly used [36], obeying the following formula: poseinterpolated(t) = (1−t)×pose1+t×pose20≤t≤1 (2.1) If tis 0, the current frame will be the same as the first keyframe. If tis 1, then the current frame will match the second keyframe. Different values lead to a weighted combination of both keyframes. Other types of interpolations can also be used: a cosine interpolation function can provide acceleration and deceleration effects at the beginning and end of an animation [37] and bilinear interpolation generates a greater variety of facial expressions when four keyframes are used, instead of two [38]. Although quick and easy, interpolation falls short to achieve smooth and realistic results so it is usually paired with other techniques, such as performance-driven methods [39]. Motion Capture Animation Motion capture (MoCap) animation, also known as performance-driven animation, emerged from the difficulty of achieving life-like characters in facial animation. It is a data-driven technique since the data obtained from external sources is mapped onto a model to create animated sequences: the majority of these methods trace facial markers placed on a performer and extract the 2D or 3D positions of these markers to animate a 3D face mesh. Accurate tracking of these points is important to maintain a consistent and realistic quality of animation [1]. Performance-driven animation has been used on movies such as The Polar Express (2004) where it allowed an actor such as Tom Hanks to drive the facial expressions of several different characters. During the shoots, 80 markers were used for the body and 152 markers were used for the face [40]. Advantages of MoCap systems include the increased speed over manually crafted animations and the potential of producing more realistic facial motion. But most marker-based motion MoCap systems use between 30-160 marker on the face which work reasonably well for capturing the motion of rigid objects but is not very effective at capturing subtle movements of the face [22]. Consequently, the animator needs to spend a significant amount of time tweaking the animation to fit the desired results. This weakness encouraged the development of new markless systems [41] and facial feature tracking from video using complex models. Procedural Animation Procedural animation is used to automatically generate animation in real-time to allow more diverse series of actions that could otherwise be created using predefined animations. With this approach, objects are animated based on physical rules, often of the real world, expressed by mathematical equations. The animator specifies these rules and the initial conditions and runs the simulation to see the results. Procedural animation requires significantly more planning time than non-procedural approaches but has the advantage of easing the build and tuning stages: by changing the input parameters it is fast and easy to create new results or to modify previous work. 16 State of the Art Several procedural animation systems have been developed to provide a range of useful character behaviours [42,43]. Most of them aiming to maximize physical realism in highly dynamic tasks such as tumbling. However, in terms of behavioural movements, procedural animation for the face is not a very explored area. 2.1.1.4 Facial Standardization Broadly speaking, within the field of computer facial animation, two kinds of people can be identified: the researchers and the artists. Researchers are mostly interested in the more technical aspects of the problems, trying to track facial features in real time in unconstrained video without markers or studying intricate methods to realistic animate anatomically correct models according to physic laws. On the other side, artists are concerned with more immediate and practical tasks of producing high quality facial animations, especially for the entertainment industry, so they use the best methods possible provided that they are compatible with the software they already master. Thus, most of the facial animation methods described in scientific articles never reach the pipelines of major movies or TV productions. But even within the research community, different groups often face problems of system interoperability because they do not use the same parameters to detect, control or animate facial movements [44]. In order to bridge these gaps, significant effort has been put on describing the face with a small set of control parameters instead of defining its complete geometry. This parameterization eases the processes of controlling facial movements, acquiring information from video analysis, reconstructing a head or transferring animations between different models. Research has shown that an ideal parameterization does not exist because it is difficult to satisfy all user demands for a broad range of facial applications. Parke developed the first facial parametric model that allowed direct creation of facial deformation by defining ad hoc parameters or by deriving parameters from the structure and anatomy of the face [38]. Since then, other approaches for facial standardization have been studied. The following sections present two standards that have been used to categorize facial expressions. FACS - Facial Action Coding System The Facial Action Coding System (FACS) is the most widely and versatile method for measuring and describing facial behaviours, having become a standard to categorize the physical expressions of emotions. It was originally published by Paul Eckman and Wallace Friesen in 1978 [45] and updated in 2002, with large contributions from Joseph Hager [46]. They determined how the contraction of each facial muscle, singly and in combination with other muscles, changes the appearance of the face by examining videotapes, studying anatomy and palpating their faces. FACS parameterizes facial expressions in terms of Action Units (AU) that are the fundamental actions of individual muscles or groups of muscles like raising left eyebrow. There are 46 AUs that represent 2.1 Character Facial Animation 17 contractions or relaxation of one or more muscles. The scores for a facial expression consist of the list of AUs that produced it. Figure 2.9: FACS. Upper row: Sample single facial AUs; Lower row: Sets of AUs for basic expressions [1] Despite its popularity and simplicity, FACS has two major weaknesses [47]: first, the AUs are purely local spatial patterns but real facial motion is rarely completely localized and second, FACS does not offer temporal components to describe the motion. Other limitations of this system include inability to describe fine eye and lip motions, and the inability to describe the co-articulation effects found most commonly in speech. MPEG-4 Facial Animation MPEG-4 is an object-based multimedia compression standard that allows encoding of different audiovisual objects (AVO) in the scene independently, developed by the ISO/IEC Moving Picture Experts Group (MPEG). Its initial objective was to achieve low bit-rate video communications but its scope was later expanded to a much broader multimedia context including images, text, graphics, 3D scenes, animation and synthetic audio. The MPEG-4 Facial Animation standard [44] specifies a face model in its neutral state and a number of Feature Points (FP) that provide spatial reference to specific positions on a human face such as major muscles and bones. It also defines a set of Facial Animation Parameters (FAP), each corresponding to a particular facial action deforming the neutral face. Facial animation sequences are generated by deforming the neutral face model according to some specific FAP values at each time instant. The standard defines 84 FPs, 66 low-level FAPs and 2 high-level FAPs, visemes and expressions. Viseme is the visual counterpart of phonemes in speech while facial expressions consist of a set of 6 basic emotions: anger, joy, sadness, surprise, disgust and fear. The low-level FAPs are based on the study of minimal facial actions and are closely related to muscle actions, precisely specifying how much a given FPs should move. All low-level FAPs are expressed in terms of the Face Animation Parameter Units (FAPUs) which are fractions of key facial features, such as the 18 State of the Art distance between the eyes. FAPUs scale FAPs for fitting any face model, allowing their interpretation in a consistent way. By coding a face model using FPs and FAPUs, developers can exchange face models without concerns about calibrations [7]. Figure 2.10: MPEG-4 Facial Animation. Left: Some Feature Points (FP); Right: A face model in its neutral state and the FPs used to define FAPUs. Fractions of distances between the marked FPs are used to define FAPU [7]. 2.2 Sketching Interaction The way humans interact with computers has evolved at the same pace as the machines themselves. Today, the barriers between users and devices are fading away with the development of new types of interfaces. As touch-screen devices become more common, this technology can provide an accessible and natural interface for sketches - rapidly executed freehand drawings - to be used in the modeling and animation processes as well as in collaborative systems. 2.2.1 Interactive Devices Human-Computer Interaction (HCI), originally known as man-machine interaction, is a term known since the early 1980s [48]. It examines the importance of usability and user-oriented interface design meaning that if focus on improving interaction between users and computing devices according to the needs and capabilities of both. Since the first digital computers were programmed, using mechanical switches and plug boards, the ways in which people interact with computers have evolved significantly and as more and more uses for technology came into play, more and more types of interfaces were created to help bridge the barrier between man and computer. A common perspective is that interfaces have passed through three loosely defined phases: command-line interfaces (CLI); graphical user interfaces (GUI) and, more recently, natural user interfaces (NUI). 2.2 Sketching Interaction 19 Figure 2.11: The evolution of interfaces. Command-line interface was the first means of human-computer interaction and the more effective way to control computing devices around the 1960s. It consisted in a kind of interactive dialogue in the form of successive lines of text (command lines) that were understood by both users and computers. The interface was usually implemented with a command line shell, which is a program that accepts commands as text input and converts them to appropriate system functions. Today, CLIs are less used by casual users but often preferred by advanced ones, since they generally provide a more concise and powerful means to control a program or operating system. With the development of ultra large scale integrated circuits, high-resolution displays and the appearance of the mouse, interfaces turned to a graphical approach. GUIs are based on metaphors, like the desktop metaphor (so called because windows are allowed to overlap, like papers on top of a desk) and rely on a known set of user interface elements, commonly referred as WIMP (Window, Icon, Menu, Pointer) [49]. 2.2.1.1 Natural User Interfaces Today, a new trend is gaining strength: NUI is an emerging paradigm shift that is reducing even more the barriers between users and machines. The term is used to refer a user interface that is invisible or becomes invisible to its users with successive learned interactions. The word "natural" is used because NUIs aim to enable users to interact with computers in the same way as they interact with the world. They rely on the ability of a user to perform relatively natural movements or gestures that they quickly discover to control the computer applications or manipulate the on-screen content. Thus, novice users quickly progress to experts. NUIs take advantage of the power of a much wider range of communication modalities, focusing on human abilities such as touch, speech, handwriting, eye-gazing, motion and higher cognitive functions such as expression, creativity, exploration, as well as combinations of them, forming multimodal interfaces. 20 State of the Art But the design and development of the technology that support new NUIs is both conceptually and practically very challenging and may require both novel software and hardware to allow input from multiple and varied sources [50]. Traditional development environments, such as Microsoft Visual Studio/.NET, Adobe Flash or Java, fall short of supporting uncommon input devices as well as handling multi-user applications, for multi-touch interaction and collaborative work. To overcome these issues, over the last few years a broad variety of heterogeneous and very specialized toolkits and frameworks have appeared like Microsoft Surface SDK [51], NUIGroup Touchlib [52], a library for creating multi-touch interaction surfaces or GlovePIE [53], that originally started as a system for emulating joystick and keyboard input using a virtual reality glove peripheral, but now supports many input devices. Few development environments that address the new requirements are available, supporting novel input devices such as physical turntables, mixing desks, multi-touch surfaces and simple vision tracking. Two examples are MAX/MSP [54] and vvvv [55], which are graphical development environments for music and video synthesis that are widely used by artist to create interactive multimedia installations. The following sections present examples of devices commonly referred as having NUIs. 2.2.1.2 Multi-touch interfaces Multi-touch devices consist of a sensing surface, like a trackpad or a touch-screen, as well as software that recognizes two or more points of contact with the surface. This plural-point awareness is often used to implement functionalities such as pinch to zoom or activating predefined programs. Multi-touch technologies have a long history. The first documented multi-touch system was developed in 1982 by the University of Toronto’s Input Research Group [56]. It consisted of a frosted-glass panel with particular optical properties so that finger pressure produced variable size black spots on an otherwise white background. Using a camera, simple image processing allowed multi-touch input. In recent years the market witnessed a proliferation of multiple finger tracking products [57], including many tablets, smartphones and digital tables. Bill Buxton presented a good overview of the evolution of the multi-touch technology [58] since its early beginnings. The following sections present some of the most relevant new trends and devices. Multi-touch Tables Multi-touch digital tables are tabletop displays that present the characteristics of multi-touch technology: a touch-screen and software to analyse the contact points. It is an innovative user-friendly technology offered in nearly any shape or size to suit any requirement. Depending on its specifications, digital tables may allow users to interact with multimedia content the same way they have interacted with physical objects using their hands, normal gestures or by putting real-world objects on the table. 2.2 Sketching Interaction 21 DiamondTouch DiamondTouch [59] is a multi-user touch technology for tabletop front-projected displays that supports small group collaboration, originally developed at Mitsubishi Electric Research Laboratories (MERL) in 2001 [60] and later licensed to Circle Twelve Inc., in 2008. It enables several different people to use the same touch-surface simultaneously without interfering with each other, or being affected by foreign objects left on the surface but its most innovative feature is the ability to identify which person is touching where. By distinguishing between different users, the system can track a person’s input and behave appropriately, controlling their access to certain functions. SMART Table The SMART Table is the first multi-touch, multi-user interactive learning center designed to stimulate collaboration among primary students. It works with both Mac and Windows operating systems and is easy-to-clean, scratch-resistant and highly durable in order to suit the needs of its young users. The SMART Table can be used together with other SMART hardware and comes with an initial set of eight learning applications but many others are available for download. The well-known 230i [61] model, with a blue top child appealing look, presented a 27 inches multi-touch display supporting up to 6 users at one time. The new SMART Table 442i [62] features a 42 inches (106.68 cm) surface with high-definition 1080p LCD display, supporting up to 40 simultaneous touches which enable up to eight students to interact simultaneously and actively collaborate to achieve shared learning goals. Microsoft PixelSense Microsoft PixelSense [63], formerly called Microsoft Surface, is an interactive surface computing platform that recognizes fingers, hands and objects placed on the screen to create a natural user interface. It allows one or more people to use touch and real world objects and share digital content at the same time. Microsoft Surface 1.0, the first version of PixelSense, was announced on May, 2007 and could recognize 52 simultaneous multi-touch points of contact in a 30 inches (76 cm) 4:3 rear projection display (1024x768). Sales of Microsoft Surface 1.0 were discontinued in 2011 in anticipation of the release of the Samsung SUR40 for Microsoft Surface and the Microsoft Surface 2.0 software platform. The current version of PixelSense, the Samsung SUR40 for Microsoft Surface was announced in 2011, and presents a 40 inches (102 cm) 16:9 LED backlit LCD display (1920x1080) with integrated PC and PixelSense technology. Tablets A tablet computer, or simply tablet, is a one-piece mobile computer, mainly controlled by touchscreen via finger gestures or a virtual keyboard, removing the need for physical input hardware components. The first commercial tablets appeared at the end of the 20th century and today many models with different sizes and features are available on the market. Tablets are lightweight and 22 State of the Art easier to carry than laptops while offer many of the same Web browsing capabilities. They are usually used for consuming multimedia content, like movies, music and books, rather than for creating content. Apple iPad The iPad is a line of tablet computers designed and marketed by Apple first released on April, 2010. The wide range of capabilities and increased usability, battery life, simplicity and overall quality of the first model, in comparison with competitor devices, earned the iPad positive reviews which defined a new standard, revolutionizing the tablet industry. However, some aspects, such as the lack of support for the Adobe Flash format were criticized. All devices run on Apple’s iOS operating system, have built-in Wi-Fi and, some models present cellular connectivity up to LTE. The most recent models, the new iPad and the iPad Mini were released on November, 2012. The new iPad [64] has a 9.7 inch LED-backlit multitouch retina display with 2048x1536 resolution at 264 pixels per inch (ppi). Microsoft Surface Microsoft Surface was the first name of the interactive surface computing platform now known as Microsoft PixelSense. Today, this name is associated to a series of tablets designed and marketed by Microsoft [65]. The Surface debuted in two models, marketed as Surface and Surface Pro: the first one with Windows RT (a special Microsoft Windows operating system designed to run on mobile devices utilizing the ARM architecture) and the other with Windows 8. Both tablets have high-definition 10.6 inches (27 cm) screens with an antifingerprint coating and 16:9 aspect ratio but have different resolutions: 1366x768 pixels for the Windows RT model and 1920x1080 pixels for the Pro model. One of the most useful features of Micrsoft Surface is the built in kickstand on the back of the unit that enables the device to maintain an upright position and become hands-free. Google Nexus Google Nexus is a line of mobile devices using the Android operative system produced by Google along with an original equipment manufacturer (OEM) partner that, today, includes smartphones and tablets (it also included a media-streaming entertainment device, Nexus Q, that was unofficial dropped due to bad reviews). The first tablet of the series, the Nexus 7 [66], developed in conjunction with Asus was unveiled in June, 2012 and shipping started the following month, being the first device to run Android version 4.1, nicknamed "Jelly Bean". Even though this is one of the smaller tablets on the market, it includes an HD touch-screen 7 inch (18 cm) display with 1280x800 pixel resolution. The second tablet of the series, the Nexus 10 [67], was manufactured by Samsung and first released in November, 2012. It runs Android 4.2 ("Jelly Bean") operative system and features a 10.1 inch display with 16:10 aspect ratio and 2560x1600 pixel resolution (300ppi), which in 2012, made it the world’s highest resolution tablet display. 2.2 Sketching Interaction 23 2.2.1.3 Spatial Interaction Multi-touch technology can enable natural user interfaces. However, most UI toolkits used to construct interfaces with such technology are traditional GUIs that fail to achieve the "transparency" desired for NUIs: our attention is projected to the multi-touch screen that remains distinguishable from the background. This idea of integrating computers seamlessly into the world is not new. In 1991, Mark Weiser, of Xerox PARC, published an article that outlined a vision of the next generation of computation where he described a model of Ubiquitous Computing, or UbiComp, where technologies "weave themselves into the fabric of everyday life until they are indistinguishable from it" [68]. At the time, there where no appropriate display systems that could work with the full diversity of input and output forms required for this UbiComp approach. Today, we are closer than ever to achieve it, with an emerging type of user interfaces that allow users to interact with computing devices in entirely new ways, such as through the motion of objects and bodies, demoting multi-touch technology to an old draft of natural user interfaces. Kinect Kinect [69] is a motion sensing input device by Microsoft for the Xbox 360 video game console and Windows PCs. It was first announced in June, 2010 under the code name "Project Natal" and released in North America on November of the same year. Kinect competed with the Wii Remote Plus and PlayStation Move motion sensor controllers but had a major advantage: this webcam-style add-on peripheral enabled users to control and interact with the Xbox 360 without the need to actually touch a game controller, using gestures and spoken commands. The device features an RGB camera, depth sensor and multi-array microphone that allow full-body 3D motion capture, facial recognition and voice recognition capabilities, changing the way people play games, watch TV and listen to music. On June, 2011 Microsoft released the Kinect SDK, allowing developers to write Kinectbased applications in C++, C# or Visual Basic .NET. Leap Motion Leap Motion [70] is a breakthrough technology focused on bringing motion control to the desktop through a small USB peripheral. The inspiration for this new technology came to its creators from the frustration surrounding 3D modeling using a mouse and keyboard versus the simplicity of molding clay in the real world. The Leap Motion controller is designed to rest on a desk in front of a monitor, creating an invisible 3D roughly 1.2 metre-square interaction space inside which the device is advertised to track hands and fingers as well as tools such as pens, pencils, and chopsticks with very high accuracy. Like Microsoft’s Kinect, the peripheral tracks human body gestures, and translates this movement into corresponding motions on a video display. According to David Holz and Michael Buckwald, co-founders of the startup Leap Motion, its input device is accurate to within 1/100 of a millimetre and 200 times more sensitive than existing motion-sensing technologies such as Microsoft’s 30 SketchFACE: Overview of the Proposed Solution 3.2 SketchFACE Approach It is the aim of this dissertation to start from a novel version of a facial sketching interface control system created by Miranda et al. [13], that allows to manipulate a model’s rig through free-hand drawing, and deploy it for Apple’s iPad. This takes advantage of the capabilities of mobile devices and the new ways of human-computer interaction enabled by them and allows to evaluate if a multi-touch capable interface is useful to decrease user’s learning curve when using software for facial modeling and animation. It is also a purpose of this work to create a simple and intuitive facial animation application that can be used in a collaborative environment with multiple users and external devices that display content in a more immersive way. This new application, inspired by the LIFEisGAME project, is called SketchFACE as it provides a Sketch-based Facial Animation Collaborative Environment. The application fits within the presented project, following the same technological challenges and interface styles and adding two additional game modes: one that explores a different interaction model to control the 3D avatar, based on 2D control canvases, and other that explores the possibility of a collaborative modeling and animation environment where multiple users can, at the same time, deform a model in a joint effort to produce an animated sequence of poses. Thus, the major results expected for the end of the dissertation are: •Deployment of a simple and intuitive sketch-based modeling and animation application for mobile devices, more specifically, the Apple’s iPad that takes advantage of the touch capabilities of tablets to implement an hybrid sketching/dragging rigging control system; •Possibility of integration of multiple players and visualization devices in order to create a collaborative and immersive environment that allows simultaneous alteration of a single model to increase challenge and entertainment; •Expansion of the scope of the project LIFEisGAME also to individuals without ASD; •Seamless integration of the new application into the LIFEisGAME prototype as additional game modes. 3.3 Challenges This research supports the creation of a simple and intuitive application for facial animation that explores a new interface: mobile devices. It also intends to evaluate if these handheld, portable, touch-capable and easily wirelessly interconnectable devices are useful to ease the task of giving life to 3D characters, thus opening new opportunities for non experienced users. However, many challenges are expected being the first ones related with the device chosen for prototyping purposes - Apple’s iPad: 3.3 Challenges 31 •The iPad’s Closed System: iPad’s operating system, iOS, is extremely restrictive. Apple is not very permissive in terms of what contents can be transferred to the iPad and creating applications for the tablet requires licenses, the signature of a a non-disclosure agreement and the payment a developer subscription. All these restrictions may prove an obstacle to new developers, that sometimes may find it very hard to find a solution for their problems that is "Apple approved". •Reduced public access: tablet technology is not yet as widespread as traditional computers. In fact, according to data from 2012 from OberCom ("Observatório da Comunicação" - The Observatory for the Media) about the Internet in Portugal, half (50,5%) of the Portuguese adults has laptop computers and desktop computers ownership rate is 35,2%. On the other hand, devices for mobile internet access (21,7%) and tablets (1,5%) are among the types of equipments held by a limited segment of the population [85]. However, mobile devices’ ownership rates tend to significantly increase, worldwide, in the near future. Forecasts by the networking giant Cisco say that the fastest growth in device adoption over the next five years will be for tablets, with a prediction of an average of 46% growth year on year, and data growth of 113% annually. This way, mobile-connected tablets will generate more traffic in 2017 (1.3 exabytes per month) that the entire global mobile network in 2012 (885 petabytes per month) [86]. At a more general level, using the multi-touch displays of mobile devices as interface for this facial animation application raises problems at two different levels, that are discussed in the following sections, mostly based on the observations of Bill Buxton’s [58]: •Disadvantages of multi-touch interfaces. •Interface design concerns. 3.3.1 Disadvantages of multi-touch interfaces Multi-touch systems have an extraordinary ability of adaptation to almost any imaginable problematic situation. As screen content can be freely modified, being, for example, able to simulate input devices (such as keyboards), touch-screens are very flexible user interfaces that enable the creation of intuitive applications, when correctly designed [87]. But taking advantage of this types of interfaces may involve considering not only their positive aspects but also some of the disadvantages associated to these equipments: •Constant visual demand: it is virtually impossible to manipulate a touch-screen without full visual attention on the display (for example when the lights are out or when our eyes are occupied elsewhere) because there is no tactile feedback that indicates the position of each interface elements. Thus, this technology has little use for visually impaired users and may even be dangerous if used while performing other activities such as driving, unless the device in question also supports speech recognition. 32 SketchFACE: Overview of the Proposed Solution •Size requirements: the size of the multi-touch display significantly influences the types of gestures that are suited for the device and the number of fingers/hands that can be used on the surface. This means that small screens may not be appropriate to some actions like taking notes or making detailed drawings. •Sunlight: multi-touch interfaces are usually incorporated in mobile devices that, as the name indicates, are specially designed to be operated outside the traditional desktops, sometimes even while in movement. This may cause exposure to adverse usability conditions, such as bright sunlight, that heavily affect visibility. Unless the device has an outstanding reflective display, it may be unusable in some light conditions. 3.3.2 Interface design concerns Creating a simple and intuitive user interface is always a major challenge in development of every application. The interface design major focus must be on the user’s experience and interaction, allowing simple and effective accomplishment of the goals and tasks at hand. But user interface design must always take into account the specific characteristics of the hardware for which the application was design for. Although often clustered together as computing devices, mobile devices, such as smartphones or tablets, are very different from traditional desktop computers, either in terms of screen size, connectivity reliability, bandwidth, battery life, and so on. Given these many differences, it is clear that designing for mobile devices, often equipped with multi-touch interfaces, is very different that designing for the desktop. As applications must embrace the device characteristics, some aspects should be considered before starting the user interface design: •Fat finger problem: when drawing on a touch-screen (without a stylus), one sometimes notices that lines are not represented on the screen exactly on the desired spot. This issue, know as "fat finger problem", arises from the fact that a finger is usually relatively large when compared to the small touch-screen. To avoid this, interface interaction elements must have a certain minimum size in order to be individually and precisely touched by human fingers. •Occlusion problem: mobile devices are manipulated with hands, and, more particularly, with fingers that touch specific parts of the screen. As fingers are not transparent, they occlude parts of the interface and the smaller the touch-screen or the more fingers are used, the more covered the display will be. This problem can be eased with clever interface design approaches [88] or with the substitution of the finger by an appropriate stylus, that is very skinny and, therefore, does not obscure the screen. •Interface interaction principles: the types of actions performed to interact with a touch technology significantly influence users’ assessment of it. The same touch technology, on the same device, can assume a very different character depending on the nature of the touch inputs it accepts or, by other words, depending on whether the interface is designed for discrete or continuous actions. Discrete actions correspond, for example, to pushing a graphical 3.4 System Design 33 button or tapping a virtual keyboard key while continuous actions correspond to gestures, such as the lateral stroke commonly used in photo-viewing application to go to the next or previous image. Discrete actions must be accompanied by graphical cues (some kind of feedback that actually works forward and not backwards) that indicate to the users that they can be performed. Some continuous actions share these properties but many do not like the "pinching" gesture to zoom in. Somehow, users must already know that it is possible to execute it, when and where. To ease user interaction, the interaction model chosen must follow known guidelines defined for the device in question. – iOS interaction model: iOS devices (such the iPad used in the case of this dissertation), are operated with specific sets of standard gestures. People become familiar with them because built-in applications use them consistently. So, when operating an iOS device with other application users are already expecting to perform successfully some particular actions to manipulate on-screen content, such as flick or drag to scroll a long list, pinch to zoom in on an image or tap a button to activate it. So, iOS developers must take into account the standards and paradigms people are comfortable with, maintaining a certain consistency encourage by Apple in its iOS Human Interface Guidelines [89]. 3.4 System Design This section intends to capture and convey the design decisions which have been made prior to the system’s deployment that served as guidelines for the implementation work. It provides a comprehensive overview of all the system’s features depicted by an interface prototype. 3.4.1 System Scenarios The system aims for seamless interaction between many users and devices. It includes two modes: a standalone version, for a single user and a collaborative environment that allows multiple simultaneous users. Each mode supports not only users using individual tablets but also supports the connection of external devices for better content visualization. Each type of device provides a different interaction model due to the capabilities it offers in result of its hardware and software (for example, a TV does not allow multi-touch input as a tablet but a tablet cannot offer a big screen for content visualization). With this in mind, it was decided that the iPad’s would support most of the direct interaction with the user (corresponding to the primarily means of input) while other devices (TV, computer screens, projectors) would provide visual feedback for the actions executed on the tablet. This option transforms the iPad into a controller for the content displayed in the external screen. External displays must be connected to the iPad via Apple TV. The different scenarios of the application are summarized in figure 3.1 and detailed below. 34 SketchFACE: Overview of the Proposed Solution Figure 3.1: Different usage scenarios for the application •Scenario 1: a single user runs the application in the iPad without any external display. Both the controls and results of the modeling and animation processes are displayed in the tablet’s screen. Being the simplest scenario, and the first to be implemented, it raises the initial problems related to sketching control method chosen, that is the foundation of the entire application. In order to create a simple sketch-based modeling and animation application that allows 3D model deformation controlled by a 2D interface, Miranda et al.’s approach can be adopted: several rig controls can be manipulated at the same time with a single stroke. Input information acquired in the 2D control interface is mapped to the 3D world coordinate system, resulting in the creation of different character poses. •Scenario 2: a single user connects an external display to the iPad. A new icon indicates the availability of the new display and, if the user chooses to use it, the iPad is used to control the 3D model displayed on the external device. This scenario adds the problem of multi-screen support, as the iPad must act as a control interface while the results of the modeling and animation processes are shown in a larger display connected to it. This obstacle can be overcome by a new functionality available in the platform chosen for system prototyping (Unity3D game engine), that allows the communication between the iPad and an external screen (of an HD TV, for example) via the digital 3.4 System Design 35 media receiver Apple TV, through AirPlay. •Scenario 3: multiple users, in the same room or not, run the application in their own iPads in multiplayer mode and cooperate in the modeling and animation processes. Scenario 3 presents the new obstacles of adding networking capabilities to the application, in order to allow cooperative work, and the necessity to define some rules that manage this collaborative intervention, since the actions of one user must not conflict with the actions of the others. Thus, a client-server approach can be used to guarantee synchronization between all the participants. Conflicting situations on the modeling process when several users are simultaneously altering a single character were avoided with the definition of an hierarchy or alteration powers. •Scenario 4: multiple users, in the same room, run the application in multiplayer mode but, as one of the iPads is connected to an external display, all of them can see the results of their cooperative work in a larger display. This scenario encompasses all the problems raised by scenarios 1, 2 and 3. 3.4.2 System Use Cases The user of SketchFACE can perform the following actions: 1. Activate external display: if an external screen is detected, a button allows the user to activate it so it begins to display a portion of the visual content of the application; 2. Deactivate external display: an active external display can be deactivated so all of the visual content is shown again on the iPad screen; 3. Create different facial poses: This can be done on-the-fly by drawing strokes on a separate 2D control canvas. Depending on how a stroke is drawn, it may deform a section of the face or drag certain control points in a determined direction. This can be also done individually or in a collaborative effort with other users; 4. Save created pose to timeline: poses in the timeline act as key-frames for the animated sequence. A created pose can be saved to the next available position of the timeline; 5. Drag timeline: When the timeline exceeds the screen size, it can be dragged back and forward to allow the access to a specific frame; 6. Animate the sequence of poses in the timeline: a "Play" button starts the animation of the poses in the timeline; 7. Pause timeline animation: a "Pause" button interrupts the animation of the currently playing sequence of poses in the timeline. This button only appears when a sequence is playing. A paused sequence will start from the same point where it stopped once the "Play" button is selected again; 36 SketchFACE: Overview of the Proposed Solution 8. Stop timeline animation: a "Stop" button allows the interruption of the animation sequence. The stopping point is not recorded so once the "Play" button is selected again, the animation will start from the first frame of the timeline; 9. Modify order of the poses in the timeline: a timeline frame can be selected and dragged to a different position; 10. Modify length of a pose in the timeline: timeline frames can be resized so a specific pose lasts more or less in the animated sequence; 11. Duplicate frame at the end of the timeline: a frame dragged and dropped at the end of the timeline will be duplicated on this new position; 12. Duplicate frame at the beginning of the timeline: a frame dragged and dropped at the beginning of the timeline will be duplicated on this new position; 13. Delete pose from timeline: a specific timeline frame dragged to the "Garbage" icon will be permanently deleted; 14. Delete all poses from timeline: a "Delete All" button will empty the timeline; 15. Clear all deformations from the 3D model: a "Clear" button resets the 3D model to its initial state, without deformations; 16. Reset a single canvas: a specific stroke drawn on a 2D canvas to deform model’s geometry can be deleted with an "Eraser" button. This will reset that canvas to its original state. 3.4.3 Interface Prototype SketchFACE aims to be a simple and intuitive sketch-based facial modeling and animation aplication. Its major feature intends to be the translation of information acquired through a 2D interface (that corresponds to strokes drawn by the users) into 3D world coordinates that deform a model’s pose. In order to fulfil this goal, while keeping special focus on simplicity, interface design must be a carefully thought issue. Interfaces must, on one side, ensure intuitiveness of the overall system so that the user requires no previous learning time or tutorials to explore all the functionalities available. On the other side, the application must follow the generic guidelines of mobile devices programming, that take maximum advantage of the hardware and software capabilities, and more particularly, follow the specific iOS user experience paradigm that corresponds to some expected interaction and navigation methods usually associated to Apple’s devices. This way, the problems and concerns discussed in sections 3.3.1 and 3.3.2 were taken into account when thinking about interface elements positioning. Figure 3.2 presents a SketchFACE interface prototype, followed by the explanation of each of its sections. 3.4 System Design 37 Figure 3.2: SketchFACE interface prototype. •Red area: 2D control canvases over a generic face image where the user draws the strokes that deform the correspondent 3D model parts. This section occupies most of the available screen area because it is specially intended to handle the majority of user’s interactions that require great precision. Some actions, like pushing an interface button do not require great precision but the actions performed in the sketching canvases, like moving a control point to refine a particular pose, are very sensitive. By making this area as big as possible, the fat finger problem is minimized, guaranteeing that input drawing areas have a certain minimum size that is compatible with their function. This area processes the events related to Use Case 3. •Orange area: buttons associated with the modeling process: "reset pose" and "reset a single canvas". Were placed next to canvases area because they help the modeling process allowed by stroke drawing performed on the canvases. This area processes the events related to Use Cases 15 and 16. •Green area: section with the visualization of the final 3D model where users can see results of the deformations, generated by the strokes drawn on the 2D control canvases, and the animated sequence. A button in this area will be available when an external display is detected. If the button is selected, the 3D model will be displayed also on the external screen. 38 SketchFACE: Overview of the Proposed Solution Results area was placed on the left part of the screen to counteract occlusion problems that arise when the user is drawing strokes, with the finger, in the canvases areas. As only an average 10-15% of the population is left-handed [90], it is more likely that users draw strokes and handle other interface controls with the right hand, thus obscuring that side of the screen. Placing the results area in the left side guarantees that, for the majority of the population, they will be visible while handling modeling interactions. This area processes the events related to Use Cases 1 and 2. •Light blue area: timeline for the keyframe poses that will generate the animated sequence. Timeline area is relatively thin in order to leave the majority of screen area to the drawing canvases. A timeline filled with poses can be dragged to the left and to the right, in order to show new available positions, with the familiar iOS scroll action. This area processes the events related to Use Cases 5, 9, 10, 11 and 12. •Dark blue area: control buttons associated to the timeline. As they correspond to actions related to the timeline, they were placed near it. This area processes the events related to Use Cases 4, 6, 7, 8, 13 and 14. Chapter 4 SketchFACE: Implementation SketchFACE is a new application specially oriented for Apple’s iPad that proposes the use of a hybrid sketching and dragging approach to manipulate a model’s control structure, known as rig, in a collaborative environment where users can join efforts to simultaneously deform and animate a 3D character. Taking advantage of the “natural interface” offered by this mobile device, it tries to create new opportunities for non-expert users and wide the target audience of animation systems while exploring the viability of collaborative work in the modeling and animation processes. The present chapter describes the implementation of the SketchFACE application. By the end of it, the reader should have a clear idea about the main design and architectural decisions made and how the different modules were implemented. 4.1 System Implementation Figure 4.1: SketchFACE interface. 39 46 SketchFACE: Implementation Model coordinates is the coordinate system local to a specific object. The origin of the object model is usually picked somewhere on the model itself but it does not have to be. The maximum control point translation transmitted by the vector (1,1) will induce the maximum joint translation/rotation allowed by the XML file. Intermediate values will lead to proportional translations/rotations. This explains how the xand ycoordinates of the world coordinate system are calculated based on the information acquired from the drawing canvases. However, the world coordinate system is composed by three coordinates: x, y and z. The zcoordinate is calculated through raycasting techniques similar to the ones described in Miranda et al.’s work [13]. In order to constraint the joint movement to the 3D mesh, the z coordinate chosen for the final position of the joint is always tangent to the model’s surface. To compute this, first a normal vector to the mesh is added to the point obtained with the mapping of the x and y coordinates, resulting in an auxiliary point, in from of the mesh. Then, the method casts a ray from the auxiliary point in the inverse normal direction and the intersection between that ray and the mesh is chosen as final position for the joint with xyz coordinates computed. This concludes the explanation about how the sketch-based method for facial modeling was implemented in SketchFACE. 4.1.3 Module 2: Collaborative Environment In order to allow multiple users to deform a single model, some networking principles needed to be implemented to create a collaborative environment. Real-time networking is a complex field but Unity3D makes it easy to add networking features to a game. However, it is useful to have some idea of the scope of networking before using it in a game. This section explains the fundamentals of networking applied to the game field, along with the specifics of Unity3D’s implementation, details the architectural scenarios evaluated and summarizes the decisions made on the topic of multiplayer support on SketchFACE. 4.1.3.1 Networking Concept Networking is communication between two or more computers, usually following a model of client-server. The client is the computer that requests the information while the server is the computer responding to the request. The server can either be a dedicated host machine used by all clients, or simply a player machine running the game (client) but also acting as the server for other players. Once a server has been established and a client has connected to it, the two computers can exchange data as demanded by gameplay. 4.1.3.2 Networking Approaches There are two common approaches for structuring a network game which are known as Authoritative Server and Non-Authoritative Server. Both approaches rely on a server connecting clients and 4.1 System Implementation 47 passing information between them and both offer privacy for end users since clients never actually connect directly with each other or have their IP addresses revealed to other clients. •Authoritative Server: this approach requires the server to perform all world simulation, application of game rules and processing of input from the player clients. Each client sends their input (in the form of keystrokes or requested actions) to the server and continuously receives the current state of the game from the server. The client never makes any changes to the game state itself. Instead, it tells the server what it wants to do, and the server then handles the request and replies to the client to explain what happened as a result. •Non-Authoritative Server: does not control the outcome of every user input. The clients themselves process user input and game logic locally, then send the result of any determined actions to the server. The server then synchronizes all actions with the world state. This is easier to implement from a design perspective, as the server really just relays messages between the clients and does no extra processing beyond what the clients do. After covering the basic architectures of networked games, it is important to explore the lowerlevels of how clients and servers can talk to each other. There are two relevant methods: Remote Procedure Calls and State Synchronization. It is not uncommon to use both methods at different points in any particular game. •State Synchronization: is used to share data that is constantly changing. The best example of this would be a player’s position in an action game. The player is always moving, running around, jumping, etc. All the other players on the network, even the ones that are not controlling this player locally, need to know where he is and what he is doing. By constantly relaying data about this player’s position, the game can accurately represent that position to the other players. This kind of data is regularly and frequently sent across the network. Since this data is timesensitive, and it requires time to travel across the network from one machine to another, it is important to reduce the amount of data that is sent as much as possible. In simpler terms, state synchronization naturally requires a lot of bandwidth, so you should aim to use as little bandwidth as possible. •Remote Procedure Calls (RPCs): are used to invoke functions on other computers across the network, although the "network" can also mean the message channel between the client and server when they are both running on the same computer. Clients can send RPCs to the server, and the server can send RPCs to one or more clients. Most commonly, they are used for actions that happen infrequently. For example, if a client flips a switch to open a door, it can send an RPC to the server telling it that the door has been opened. The server can then send another RPC to all clients, invoking their local functions to open that same door. They are used for managing and executing individual events. An RPC call can have as many parameters as desired but numerous or large parameters will 48 SketchFACE: Implementation have an impact on the network bandwidth involved. Unlike a normal function call, an RPC needs an additional parameter to denote the recipients of the request which can be: only the server, everyone or everyone except the sender. RPC calls can also be buffered. Buffered RPC calls are stored up and executed in the order they were issued for each new client that connects. This can be a useful way to ensure that a late coming player gets all necessary information to start. 4.1.3.3 Master Server Each "game room" where players meet requires one server that assures communications between them. The server can be a machine that is also running the game, which mean that one of the players may act as server and the other players connected to it act as clients, or it can be a dedicated machine that is not participating in the game. Either case, it is always necessary one Master Server used to matchmake servers and clients so servers can be advertised and compatible clients can connect to running games. As such, the Master Server is essentially a basic lobby server, but with some special features. Its purpose is also to hide IP address and port details, and to perform technical tasks around setting up network connections which otherwise would be impossible, like acting as a proxy server to let clients behind a firewall act as a game server. Unity Technologies has a fully deployed online Master Server available for testing purposes. It is a built-in meeting place for games that are actively seeking clients, and player clients who want to connect to them. Unfortunately, it has no uptime guaranty which makes it an unreliable solution. However, the source code of Unity’s Master Server is freely available for anyone to use and the server can be deployed on Windows, Linux and Mac OS [92]. It can also be customized with modifications to the source code. 4.1.3.4 SketchFACE Networking Approach The chosen networking approach for the collaborative environment of the SketchFACE application relied on a Master Server built and running from the source code made available at Unity3D website since it fulfils all the necessary networking requirements and overcomes the online server unreliability. The communication between users is assured by buffered RPC calls, since the data exchanged is not constantly changing, in order to ensure model, canvas and timeline synchronism between all users, in the following situations: •Lock canvas: when a user starts using a particular canvas by beginning a stroke or moving a control point (Use Case 3), it must be locked for all other users to avoid conflicting model deformations; •Unlock canvas: when a user ends the interaction with a particular canvas by ending the stroke or releasing a control point (end of Use Case 3), it must be unlocked so other user can use it; 4.1 System Implementation 49 •Send pose information: when the model’s pose is altered by any user (end of Use Case 3), this information must be send to other users so they can recreate the pose in their devices and thus, maintain the synchronism inside the collaborative environment; •Reset pose: when one of the users selects the button responsible for making the model return to the initial neutral neutral pose (Use Case 15) this transformation must be propagated to all other users; •Reset a single canvas: sends the information to reset a particular canvas (control points and reference curve) and the 3D model bones associated, to all the connected users, when one of them performs this action (Use Case 16); •Save pose to timeline: propagates to all the connected users a new frame saved to the timeline (Use Case 4); •Delete timeline frame: when a user deletes one of the timeline frames (Use Case 13), it disappears from the timeline of all the connected users; •Drop frame at the end of the timeline: a frame dragged and dropped at the end of the timeline will be duplicated on this new position (Use Case 11). If a user performs this action it will be replicated on the timeline of all other users; •Drop frame at the beginning of the timeline: this action is similar to the previous one but occurs when a frame is dragged and dropped at the beginning of the timeline (Use Case 12); •Reposition frame: when a user drags and drops a frame to a new timeline position (Use Case 9) it must be repositioned for all other users; •Change frame length: when a user changes the length (that corresponds to the duration) of a frame (Use Case 10), this transformation must be propagated to all other users; •Clear timeline: if one of the users clears the entire timeline, by selecting the correspondent button (Use Case 14), all frames from the timelines of all other connected users will be deleted. 4.1.4 Module 3: Content Visualization Multiple screen support was always a very desired feature among Unity game developers. Connecting external screens supporting Full HD to iOS devices is an important part of creating immersive user experience in native games and multimedia applications. iOS SDK makes supporting this feature really easy for native applications but with Unity3D engine this is was not an easy task. By the beginning of the development of this dissertation Unity3D did not support rendering to multiple targets which meant it only could create one rendering window. And if we wanted an iPad to be one of the visualization displays, the problem got even more complicated because there 50 SketchFACE: Implementation were some options available to mirror the iPad screen to an external display but span it so we could show a different portion of Unity’s rendering scene in each display was extremely tricky. To overcome this problem some solutions were analysed: •Render a single really-wide image: the goal was to create a scene with a custom resolution that corresponded to the sum of the resolution of all connected devices so the image spanned across the available displays. In order to detect the connection/disconnection of external displays this solution had to be integrated with plugins. When developing games in Unity3D it is often necessary to access platform specific features, hardware or anything not available to Unity via the API. Plugins are a way to establish a bridge between the code in Unity (C#, Javascript or Boo) and the native platform. It allows the access to any non-supported features by creating a native binary bundle which can be accessed through an interface in the Unity script. There are some examples of this kind of implementation but displays were always connected to a desktop PC and not to the iPad. A prototype of this solution was deployed but only successfully for multiple desktop screens, not involving an iPad. •Run different instances of the application on each screen and make then interact: in other words, we would have a separate executable assigned to each screen. The communication between them could be implemented via text files or via multiplayer options such as using localhost. However, none of the above solutions proved to be ideal. Fortunately, during the development of the work, a new version of Unity3D (Unity 4.1) became available and with it, new features, in particular full multi-screen support using AirPlay [93]. AirPlay is a protocol stack developed by Apple Inc. that allows wireless streaming of audio, video and photos, together with related metadata between devices. It allows to wirelessly stream what is on one iOS device to an HD TV and speakers via the digital media receiver Apple TV. This feature is only properly implemented for iOS but this was exactly the solution for the multi-screen support problem in question. A new class "Display" shows an array of connected displays and allows selective rendering for each one so it is possible to render one camera for the iPad and another to an external display connected via Apple TV. The iPad can now be used as a game controller, running and controlling the game from the iOS handheld device whilst the action unfolds on a a bigger screen. Chapter 5 Results and Validation A series of experiments was conducted in order to validate the system deployed for this dissertation. It was evaluated if this new approach based on mobile devices was more intuitive that the alternative based on the traditional computer with mouse and it was also tested the relevance and utility of the collaborative environment. This was done by computing the time and effort it took for users to create a particular facial expression in the several experimental scenarios and with questionnaires about subjective usability opinion. The target audience chosen comprised people with no previous experience with modeling and animation tools but with relatively good familiarity with electronic devices in order to validate if the system was appropriate for novice users. This chapter describes the experiments conducted, presents their results and exposes the conclusions that can be drawn from them. 5.1 Experiment Design During the testing stage, two major experiments were performed: an interface experiment and acollaborative environment experiment. The interface experiment was designed to evaluate the effectiveness of the interface device chosen to deploy the application. It intended to analyse if mobile touch-capable devices such as tablets and, in this particular case, the iPad, are more natural, intuitive and simpler to use than the traditional computer whose interaction model is based on mouse clicks. Therefore, the following research questions were formulated: •Q1: Are mobile device based approaches, with tactile interfaces, as easy and intuitive for creating facial expressions with a sketch-based method as the traditional computer/mouse based approaches? •Q2: Is the necessary time to create facial expressions using the iPad’s touch-capable interface and the computer with a mouse input device similar? The collaborative environment experiment evaluated if such an environment where multiple users can deform, at the same time, the same model, can benefit the task of creating facial expressions. Therefore, the following research question was formulated: 51 52 Results and Validation •Q3: A collaborative facial animation environment is as useful as an individual approach for the task of creating facial expressions? To collect data to answer the research questions, some objective measurements were performed: it was recorded the time and number of strokes, control point movements and canvas resets users made to create a pose. Subjective usability data was collected with a questionnaire, that can be observed in Appendix A, comprising four questions related to user background, three questions about the interface experiment, two questions about the collaborative environment experiment and two open questions about improvements suggestions and additional comments. 5.1.1 Participants Both experiments were performed with a group of 22 people with good or very good level of familiarity with electronic devices (such as computers, tablets, smartphones, etc) and none or poor experience with modeling and animation tools. The participants had an average age of 29,77 years (SD = 3,74) and relatively balanced sex distribution, with 9 males and 13 females performing the experiments as can be seen on table 5.1. Table 5.1: Information about the experiments participants. Sex Age Male Female Av SD Participants 9 13 29,77 3,74 5.1.2 Interface Experiment The interface experiment began with an introductory description of the application and its development context as well as a brief explanation of the tests to be performed. Then, two tests were conducted, one for each evaluated interface device: the computer and the iPad. Each participant performed the tests individually in sessions of about 10 to 20 minutes. The computer test was composed of two phases: •Training phase: the first contact of the participants with SketchFACE. They were given no explanation about the application and asked to explore the interface during 2 minutes. •Task Phase: An explanation about every functionality of the application was given to the participants (2 minutes) and then they were asked to reproduce a particular facial expression (Figure 5.1). There was no time limit to conclude the task and the only direction provided was that the users needed to copy, as faithfully as possible, the given expression. Then, an iPad with the application was presented to the participants. After a short explanation about the new interaction model allowed by this device (based on strokes drawn with the finger) and the specific functionalities featured on the mobile device (multi-display support), the test had also two phases: 5.2 Experiment Results 53 •Training phase: the participants had 2 minutes to explore and get familiar with the new interface and interaction model; •Task Phase: the participants were were asked to reproduce the same facial expression copied in the previous test (Figure 5.1). There was no time limit to conclude de task and the only direction provided was that the users needed to copy, as faithfully as possible, the given expression. Figure 5.1: Facial expression that the participants were asked to reproduce using the SketchFACE application, both on the computer and on the iPad. 5.1.3 Collaborative Environment Experiment The same participants of the previous experiment were now paired up. The experiment was composed by 3 tests where users were asked to reproduce the 3 facial expressions of figure 5.2, being the first one the same already copied in the previous experiment. Users had to work together, deforming at the same time the same model, in the collaborative environment provided by SketchFACE. There was no time limit for neither one of the tests and the only rule was that the user should reproduce, as faithfully as possible, the given expressions. 5.2 Experiment Results In order to answer question Q1 and to try to understand what type of devices are more intuitive as interfaces for creating facial expressions with a sketch-based method, question 2.1.1 of the questionnaire was analysed: •From the following options, choose the one that best matches your opinion: a) I consider the computer interface easier and more intuitive. 54 Results and Validation Figure 5.2: Facial expressions that the participants were asked to reproduce using the SketchFACE application, in the collaborative environment mode. b) I consider the iPad interface easier and more intuitive. The results can be seen in figure 5.3. Figure 5.3: Distribution of the answers to the question about what interface device was more intuitive. The analysis of the chart shows that the participants consider the iPad easier and more intuitive. This way, the null hypothesis is rejected and the alternative hypothesis, that there is significant difference between the level of intuitiveness of both devices is different is accepted, being the iPad considered the more intuitive interface. User’s activities, evaluated through the analysis of the experience’s log that recorded every stroke created, control point moved, complete reset made or single canvas reset performed, also support this conclusion. Although with a relatively small margin, users had to perform less actions on the iPad to achieve the same results as can be seen in figure 5.4. In order to answer question Q2 and assess the existence of significant difference between the time spent creating a facial expression on the iPad and on the computer, a difference of means test was performed. Given the relatively small number of cases evaluated (N = 22) and the fact that 5.2 Experiment Results 55 Figure 5.4: Average number of user actions during the interface experiment by category. a normality test indicated a non-normal distribution of the results, the time comparison between both device interfaces was performed with a non-parametric Mann-Whitney test whose results can be seen on table 5.2. Table 5.2: Results of the Mann-Whitney test performed. Av SD Min - Max (seconds) U p Computer 237,41 107,05 112 - 512 188,00 n.s. iPad 187,91 101,73 55 - 431 The analysis of table 5.2 indicates that the average time to create a facial pose with SketchFACE on the iPad was smaller than on the computer (as can also be seen on figure 5.5). However, time differences between both interface devices were not statistically significant (U = 188,00; p > 0,05). This way, the null hypothesis that states that there are no significant differences between the necessary time to create facial expression using the iPad’s touch-capable interface or the computer must be accepted. Regarding the collaborative environment, in order to evaluate if it made the task of creating a specific facial expression easier, question 2.2.1 from the questionnaire was analysed: •From the following options, choose the one that best matches your opinion: a) I consider that the collaborative environment eases the accomplishment of the task. b) I consider that the collaborative environment does not ease the accomplishment of the task. 62 Conclusion and Future Work •Further experiments – Larger group of participants: the performed validation experiments ended up in many non statistically significant results. The outcome could have been different if a wider number of participants had been used and if different groups would have been used for each tested scenario. Further experiments can explore these options; – Different user profiles: it could have been interesting to widen the experiments to multiple groups of people with different profiles. The performed tests only targeted people with good or very good level of familiarity with electronic devices, none or poor experience with modeling and animation tools and average age of 30 years. Using also participants with different age groups or with prior experience with animation tools can add value to the results obtained from these experiments. – Further evaluation of the collaborative environment: although the system was implemented to support more than two simultaneous users, no tests were conducted in order to validate other scenarios with more participants. As some aspects of the collaborative environment need further testing, future experiments can address this topic. Appendix A Validation Experiments Documents A.1 Usability Questionnaire The following pages show the usability questionnaire presented to the participants of the validation experiments performed on SketchFACE. 63 ____________________________________________________________________________________________________ SKETCH-BASED FACIAL MODELING AND ANIMATION: AN APPROACH BASED ON MOBILE DEVICES Ana Luísa de Vila Fernandes Orvalho [email protected] | [email protected] ___________________________________________________________________________________ Orientador: Prof. Augusto Sousa (FEUP) | Co-orientadora: Prof. Verónica Orvalho (FCUP) ___________________________________________________________________________________ Introdução  O presente questionário pretende, do ponto de vista do utilizador, avaliar a usabilidade e validar a aplicação SketchFACE, desenvolvida no contexto de uma dissertação do Mestrado Integrado em Engenharia Eletrotécnica e de Computadores.  As respostas são completamente voluntárias e confidenciais. Descrição do Projeto Animação facial que satisfaça as expectativas dos espectadores é muito difícil de se alcançar. O tradicional software de modelação e animação de modelos 3D, apesar de muito poderoso, é bastante complexo e exige um tempo de aprendizagem significativo para se conseguir explorar e dominar os seus múltiplos menus e funcionalidades. Esta dissertação propõe o uso de uma abordagem baseada em desenhos para manipular a estrutura de controlo dos modelos 3D, o rig, num ambiente colaborativo, em que vários utilizadores podem unir esforços para simultaneamente deformar e animar uma personagem 3D. Foi desenvolvido um protótipo de uma nova aplicação, designada SketchFACE (Sketch-based Facial Animation Collaborative Environment), especialmente orientado para o iPad. Aproveitando a interface tátil “mais natural” deste dispositivo móvel, tentou-se criar novas oportunidades para utilizadores com pouca ou nenhuma experiência, alargando, assim, o público-alvo dos sistemas de animação enquanto se explora a viabilidade do trabalho colaborativo nos processos de modelação e animação. A solução proposta foi inspirada no projeto LIFEisGAME – um jogo sério que pretende ajudar crianças com problemas do espectro do autismo a reconhecer e expressar emoções através de expressões faciais de forma interativa e creativa. ___________________________________________________________________________________ 1) Perfil do Utilizador 1.1) Idade: _____________ Sexo: M  F  1.2) Nível de familiarização com dispositivos eletrónicos (computadores, tablets, smartphones…)  a) Nenhum  b) Pouco  c) Bom  d) Muito Bom 1.3) Nível de experiência com ferramentas de modelação e animação facial.  a) Nenhum  b) Pouco  c) Bom  d) Muito Bom ID: _____________ 1.4) Background artístico  a) Nenhum  b) Pouco  c) Bom  d) Muito Bom 2) Sobre o protótipo SketchFACE 2.1) Interface com o utilizador 2.1.1) Das seguintes opções escolha a que melhor corresponde à sua opinião:  a) Considero que a interface do computador é mais fácil e intuitiva de usar.  b) Considero que a interface do iPad é mais fácil e intuitiva de usar. 2.1.2) Considera a interface da aplicação suficientemente clara? Sim  Não  2.1.3) Se não, porquê? ________________________________________________________ ___________________________________________________________________________ ___________________________________________________________________________ 2.2) Ambiente colaborativo 2.2.1) Das seguintes opções, escolha a que melhor corresponde à sua opinião:  a) Considero que o ambiente colaborativo facilita a realização da tarefa proposta.  b) Não considero que o ambiente colaborativo facilite a realização da tarefa. 2.2.2) Das seguintes opções, escolha a que melhor corresponde à sua opinião:  a) Considero que esta ferramenta é inovadora.  b) Não considero que esta ferramenta seja inovadora. 3) Registe aqui a sua apreciação global sobre este protótipo SketchFACE 3.1) Sugestões de melhoria: ____________________________________________________ ___________________________________________________________________________ ___________________________________________________________________________ 3.2) Comentários gerais: _______________________________________________________ ___________________________________________________________________________ ___________________________________________________________________________ Data: _____ /______/_____ Obrigada pela sua participação  66 Validation Experiments Documents References [1] Zhigang Deng and Junyong Noh. Computer facial animation: A survey. In Zhigang Deng and Ulrich Neumann, editors, Data-Driven 3D Facial Animation, pages 1–28. Springer London, 2007. URL: http://dx.doi.org/10.1007/978-1-84628-907-1_1. [2] S. Villagrasa, A. Susín Sánchez, et al. Face! 3d facial animation system based on facs. 2010. [3] Rob Bredow, David Schaub, Daniel Kramer, Matthew Hausman, Danny Dimian, and R. Stirling Duguid. Surf’s up: the making of an animated documentary. In ACM SIGGRAPH 2007 courses, SIGGRAPH ’07, pages 1–123, New York, NY, USA, 2007. ACM. URL: http:// doi.acm.org/10.1145/1281500.1281600,doi:10.1145/1281500.1281600. [4] O. Alexander, M. Rogers, W. Lambeth, M. Chiang, and P. Debevec. Creating a photoreal digital actor: The digital emily project. In Visual Media Production, 2009. CVMP’09. Conference for, pages 176–187. IEEE, 2009. [5] David Komorowski, Vinod Melapudi, Darren Mortillaro, and Gene S. Lee. A hybrid approach to facial rigging. In ACM SIGGRAPH ASIA 2010 Sketches, SA ’10, pages 42:1–42:2, New York, NY, USA, 2010. ACM. URL: http://doi.acm.org/10.1145/1899950. 1899992,doi:10.1145/1899950.1899992. [6] W. GRUBB. Facial animation rig for delgo. Tutorial Notes, 12, 2009. [7] K. Balci. Xface: Mpeg-4 based open source toolkit for 3d facial animation. In Proceedings of the working conference on Advanced visual interfaces, pages 399–402. ACM, 2004. [8] B. Paulson, P. Rajan, P. Davalos, R. Gutierrez-Osuna, and T. Hammond. What!?! no rubine features?: using geometric-based features to produce normalized confidence values for sketch recognition. In HCC Workshop: Sketch Tools for Diagramming, pages 57–63, 2008. [9] LIFEisGAME. Learning of facial emotions using serious games, 2010. URL: http:// www.portointeractivecenter.org/lifeisgame/ [last accessed 2013-01-27]. [10] FCT. Fundação para a ciência e a tecnologia. URL: http://www.fct.pt/ [last accessed 2013-01-27]. [11] UT Austin|Portugal. Internationa collaboratory for emerging technologies (colab). URL: http://utaustinportugal.org/ [last accessed 2013-01-27]. [12] IT. Instituto de telecomunicações. URL: http://www.it.up.pt/ [last accessed 201301-27]. [13] José Carlos Miranda, Xenxo Alvarez, João Orvalho, Diego Gutierrez, A. Augusto Sousa, and Verónica Orvalho. Sketch express: A sketching interface for facial animation. Computers 67 68 REFERENCES & Graphics, 36(6):585 – 595, 2012. URL: http://www.sciencedirect.com/ science/article/pii/S0097849312000416,doi:10.1016/j.cag.2012.03. 002. [14] Frederick I. Parke. Computer generated animation of faces. In Proceedings of the ACM annual conference - Volume 1, ACM ’72, pages 451–457, New York, NY, USA, 1972. ACM. URL: http://doi.acm.org/10.1145/800193.569955,doi:10.1145/800193. 569955. [15] K. Ritchie, J. Callery, and K. Biri. The Art of Rigging: A definitive guide to character technical direction with Alias Maya. CG Toolkit, 2005. [16] M. Mori. The uncanny valley, translated by karl f. macdorman and takashi minato. Energy, 7(4):33–35, 1970. [17] D. Hanson, A. Olney, S. Prilliman, E. Mathews, M. Zielke, D. Hammons, R. Fernandez, and H. Stephanou. Upending the uncanny valley. In Proceedings of the National Conference on Artificial Intelligence, volume 20, page 1728. Menlo Park, CA; Cambridge, MA; London; AAAI Press; MIT Press; 1999, 2005. [18] Canetti Y. Piretti M. Schleifer J., Scaduto-Mendola R. Character setup from rig mechanics to skin deformations: A practical approach. SIGGRAPH ’02 Course Notes, 2002. [19] Nikolaos Ersotelos and Feng Dong. Building highly realistic facial modeling and animation: a survey. The Visual Computer, 24:13–30, 2008. URL: http://dx.doi.org/10.1007/ s00371-007-0175-y,doi:10.1007/s00371-007-0175-y. [20] V. Orvalho, P. Bastos, F. Parke, B. Oliveira, and X. Alvarez. A facial rigging survey. In Eurographics 2012-State of the Art Reports, pages 183–204. The Eurographics Association, 2012. [21] T. McLaughlin and S.S. Sumida. The morphology of digital creatures. In ACM SIGGRAPH, pages 05–09, 2007. [22] V.C.T. Orvalho. Reusable facial rigging and animation: Create once, use many. PhD thesis, Universitat Politècnica de Catalunya, 2007. [23] J. Fordham. Middle earth strikes back. Cinefex, 2003. [24] Pushkar Joshi, Wen C. Tien, Mathieu Desbrun, and Frédéric Pighin. Learning controls for blend shape based realistic facial animation. In ACM SIGGRAPH 2005 Courses, SIGGRAPH ’05, New York, NY, USA, 2005. ACM. URL: http://doi.acm.org/10. 1145/1198555.1198588,doi:10.1145/1198555.1198588. [25] A. Ward. Game character development with maya. New Riders, 2005. [26] N. Burtnyk and M. Wein. Interactive skeleton techniques for enhancing motion dynamics in key frame animation. Commun. ACM, 19(10):564–569, October 1976. URL: http: //doi.acm.org/10.1145/360349.360357,doi:10.1145/360349.360357. [27] J. P. Lewis, Matt Cordner, and Nickson Fong. Pose space deformation: A unified approach to shape interpolation and skeleton-driven deformation, 2000. REFERENCES 69 [28] T.W. Sederberg and S.R. Parry. Free-form deformation of solid geometric models. In ACM Siggraph Computer Graphics, volume 20, pages 151–160. ACM, 1986. [29] S. Coquillart. Extended free-form deformation: a sculpturing tool for 3D geometric modeling, volume 24. ACM, 1990. [30] Prem Kalra, Angelo Mangili, Nadia Magnenat Thalmann, and Daniel Thalmann. Simulation of facial muscle actions based on rational free form deformations. Computer Graphics Forum, 11(3):59–69, 1992. URL: http://dx.doi.org/10.1111/1467-8659. 1130059,doi:10.1111/1467-8659.1130059. [31] K. Singh and E. Fiume. Wires: a geometric deformation technique. In Proceedings of the 25th annual conference on Computer graphics and interactive techniques, pages 405–414. ACM, 1998. [32] M.A. Warburton. Physically-based facial modelling and animation with wrinkles: 12 month report. 2011. [33] Stephen M. Platt and Norman I. Badler. Animating facial expressions. SIGGRAPH Comput. Graph., 15(3):245–252, August 1981. URL: http://doi.acm.org/10.1145/ 965161.806812,doi:10.1145/965161.806812. [34] J. Osipa. Stop staring: facial modeling and animation done right. Sybex, 2010. [35] John Lasseter. Principles of traditional animation applied to 3d computer animation. In Proceedings of the 14th annual conference on Computer graphics and interactive techniques, SIGGRAPH ’87, pages 35–44, New York, NY, USA, 1987. ACM. URL: http://doi. acm.org/10.1145/37401.37407,doi:10.1145/37401.37407. [36] Frederic Pighin, Jamie Hecker, Dani Lischinski, Richard Szeliski, and David H. Salesin. Synthesizing realistic facial expressions from photographs. In ACM SIGGRAPH 2006 Courses, SIGGRAPH ’06, New York, NY, USA, 2006. ACM. URL: http://doi.acm.org/10. 1145/1185657.1185859,doi:10.1145/1185657.1185859. [37] K. Waters and T. Levergood. An automatic lip-synchronization algorithm for synthetic faces. In Proceedings of The second ACM international conference on Multimedia, pages 149–156. ACM, 1994. [38] F.I. Parke. A parametric model for human faces. Technical report, DTIC Document, 1974. [39] Jun yong Noh and Ulrich Neumann. A Survey of Facial Modeling and Animation Techniques. 1998. [40] D. Bennett. The faces of the polar express. In ACM Siggraph 2005 Courses, page 6. ACM, 2005. [41] V. Blanz, C. Basso, T. Poggio, and T. Vetter. Reanimating faces in images and video. Computer Graphics Forum, 22(3):641–650, 2003. URL: http://dx.doi.org/10.1111/ 1467-8659.t01-1-00712,doi:10.1111/1467-8659.t01-1-00712. [42] W.L. Wooten and J.K. Hodgins. Simulating leaping, tumbling, landing and balancing humans. In Robotics and Automation, 2000. Proceedings. ICRA’00. IEEE International Conference on, volume 1, pages 656–662. IEEE, 2000. 70 REFERENCES [43] I.D. Horswill. Lightweight procedural animation with believable physical interactions. Computational Intelligence and AI in Games, IEEE Transactions on, 1(1):39 –49, march 2009. doi:10.1109/TCIAIG.2009.2019631. [44] I.S. Pandzic and R. Forchheimer. Mpeg-4 facial animation. The standard, implementation and applications. Chichester, England: John Wiley&Sons, 2002. [45] P. Ekman and W. V. Friesen. Facial action coding system: a technique for the measurement of facial movement. 1978. [46] P. Ekman, W.V. Friesen, and J.C. Hager. Facial action coding system. A Human Face, 2002. [47] Irfan Essa, Sumit Basu, Trevor Darrell, and Alex Pentland. Modeling, tracking and interactive animation of faces and heads using input from video. In IN PROCEEDINGS OF COMPUTER ANIMATION CONFERENCE, pages 68–79. IEEE Computer Society Press, 1996. [48] Achim Ebert, NahumD. Gershon, and GerritC. Veer. Human-computer interaction. KI - Künstliche Intelligenz, 26:121–126, 2012. doi:10.1007/s13218-012-0174-7. [49] Daniel Wigdor and Dennis Wixon. Brave NUI World: Designing Natural User Interfaces for Touch and Gesture. Morgan Kaufmann Publishers Inc., San Francisco, CA, USA, 1st edition, 2011. [50] R. Harper. Being human: Human-computer interaction in the year 2020. Microsoft Research, 2008. [51] Microsoft. Surfice sdk. URL: http://msdn.microsoft.com/en-us/library/ ff727815(v=Surface.20).aspx [last accessed 2013-02-04]. [52] NUIGroup. Touchlib. URL: http://nuigroup.com/touchlib/ [last accessed 201302-04]. [53] Carl Kenner. Glovepie. URL: http://glovepie.org/glovepie.php [last accessed 2013-02-04]. [54] Cycling ’74. Max/msp/jitter. URL: http://www.cycling74.com/ [last accessed 201302-01]. [55] vvvv: a multipurpose toolkit. URL: http://vvvv.org/ [last accessed 2013-02-01]. [56] N. Metha. A flexible machine interface. MA Sc. Thesis, Department of Electrical Engineering, University of Toronto, 1982. [57] Daniel Wigdor, Joe Fletcher, and Gerald Morrison. Designing user interfaces for multitouch and gesture devices. In CHI ’09 Extended Abstracts on Human Factors in Computing Systems, CHI EA ’09, pages 2755–2758, New York, NY, USA, 2009. ACM. URL: http:// doi.acm.org/10.1145/1520340.1520399,doi:10.1145/1520340.1520399. [58] B. Buxton. Multi-touch systems that i have known and loved. Microsoft Research, 2007. [59] Circle Twelve Inc. Diamondtouch. URL: http://www.circletwelve.com/ products/diamondtouch.html [last accessed 2013-02-04]. REFERENCES 71 [60] P. Dietz and D. Leigh. Diamondtouch: a multi-user touch technology. In Proceedings of the 14th annual ACM symposium on User interface software and technology, pages 219–226. ACM, 2001. [61] SMART Technologies Inc. Smart table 230i. URL: http://smart.lsk.hu/edu/ tamogatas/letoltes/pdf/adatlapok/st230i.pdf [last accessed 2013-02-04]. [62] SMART Technologies Inc. Smart table 442i. URL: http:// downloads01.smarttech.com/media/sitecore/en/pdf/products/table/ mktg-913-rev05-ed-fs-table_final.pdf [last accessed 2013-02-04]. [63] Microsoft. Pixelsense. URL: http://www.microsoft.com/en-us/pixelsense/ default.aspx [last accessed 2013-02-04]. [64] Apple Inc. ipad. URL: http://www.apple.com/ipad/features/ [last accessed 2013-02-04]. [65] Microsoft. Microsoft surface. URL: http://www.microsoft.com/Surface/en-US [last accessed 2013-02-04]. [66] Google. Nexus 7. URL: http://www.google.com/nexus/7/ [last accessed 2013-0204]. [67] Google. Nexus 10. URL: http://www.google.com/nexus/10/ [last accessed 201302-04]. [68] M. Weiser. The computer for the 21st century. Scientific American, 265(3):94–104, 1991. [69] Microsoft. Kinect. URL: http://www.xbox.com/en-US/kinect [last accessed 201302-04]. [70] Leap Motion. Leap. URL: https://www.leapmotion.com/ [last accessed 2013-0204]. [71] L. Olsen, F.F. Samavati, M.C. Sousa, and J.A. Jorge. Sketch-based modeling: A survey. Computers and Graphics (Pergamon), 33(1):85–103, 2009. cited By (since 1996) 48. URL: http://www.scopus.com/inward/record.url?eid=2-s2. 0-59249095193&partnerID=40&md5=79316233b31b419e83ca721757c11dfa. [72] S. Saga. A freehand interface for computer aided drawing systems based on the fuzzy spline curve identifier. In Systems, Man and Cybernetics, 1995. Intelligent Systems for the 21st Century., IEEE International Conference on, volume 3, pages 2754 –2759 vol.3, oct 1995. doi:10.1109/ICSMC.1995.538199. [73] Tevfik Metin Sezgin, Thomas Stahovich, and Randall Davis. Sketch based interfaces: early processing for sketch understanding. In Proceedings of the 2001 workshop on Perceptive user interfaces, PUI ’01, pages 1–8, New York, NY, USA, 2001. ACM. URL: http:// doi.acm.org/10.1145/971478.971487,doi:10.1145/971478.971487. [74] Tevfik Metin Sezgin, Thomas Stahovich, and Randall Davis. Sketch based interfaces: early processing for sketch understanding. In ACM SIGGRAPH 2006 Courses, SIGGRAPH ’06, New York, NY, USA, 2006. ACM. URL: http://doi.acm.org/10.1145/1185657. 1185783,doi:10.1145/1185657.1185783.