A Methodological Investigation into the Application of the MVP Support System in Ensemble Performance Practice by Wind Instrument Players

A Methodological Investigation into the Application of the MVP Support System in Ensemble Performance Practice by Wind Instrument Players Yasumasa Yamaguchi Sendai University 2-2-18, Funaoka Minami, Shibatamachi [email protected] l ABSTRACT This study explores the application and evaluation of the MVP Support System during actual ensemble practice sessions. The MVP Support System captures the sound of performers through a microphone, estimates pitch using a machine learning model, and provides pitch feedback through a browser-based interface. Unlike conventional tuners that use needle and dial indicators, this system visualizes pitch deviations using color: red indicates a pitch higher than the reference, blue indicates a lower pitch, and green indicates correct intonation. While previous research on the MVP Support System has been conducted in controlled laboratory environments, this study applies the system in real-world ensemble rehearsal settings and discusses its methodological implications. 1. INTRODUCTION Intonation is one of the most fundamental technical and expressive elements in wind instrument performance, especially in ensemble contexts such as chamber music, wind bands, and orchestras [1]. For many student musicians, particularly those in the early stages of training, maintaining accurate pitch while simultaneously attending to musical expression and ensemble coordination remains a significant challenge. In Japanese secondary school music education, it is common practice for students to place visual tuners with dials and needles on their music stands and check pitch during performance. While this strategy may support pitch accuracy, it can negatively impact performance posture and ensemble awareness by encouraging players to divert their gaze away from the conductor and coperformers. The emergence of wearable and IoT (Internet of Things) devices—especially smart glasses—has opened new possibilities for integrating real-time feedback into musical performance. Within this context, the MVP (Musical Visual Pitch) Support System has been developed to provide performers with real-time, visually intuitive pitch feedback while minimizing interference with natural playing posture and visual attention. The aim of this study is to evaluate the practical use of the MVP Support System in a realistic ensemble rehearsal setting and to explore methodological considerations for future empirical studies in this domain. 2. OVERVIEW OF THE mvp SUPPORT SYSTEM The MVP Support System captures audio from the performer using a USB microphone and displays pitch feedback in color on a smart glasses interface via browser-based processing. It is designed specifically for use with Google’s Glass Enterprise Edition 2, enabling performers to maintain natural posture and visual attention while receiving real-time, intuitive feedback on pitch accuracy. In addition, because the system processes data through a web browser, it offers excellent scalability across a wide range of devices while maintaining a high degree of flexibility in system customization. For pitch estimation, the system employs the ml5.js machine learning library developed by NYU ITP. This library includes a "Pitch Detection" module based on the CREPE (Convolutional Recurrent Neural Network for Pitch Estimation) algorithm [2][3], which has been shown to outperform traditional FFT-based pitch detection methods in terms of accuracy and responsiveness [5]. It is also known for its robustness against harmonic interference and noise, making it well-suited for use in musical contexts. The high responsiveness of ml5.js has led to its adoption in other musictechnology applications such as browser-based score-following systems in multi-voice children's music [6] Given the limited display size of smart glasses, the system replaces traditional needle-and-dial indicators with a simple colorcoded feedback system: red for sharp pitches, blue for flat, and green for in-tune. A pitch is considered "in tune" if it falls within ±1% of the reference frequency in Hertz. As the system operates entirely in the browser, it is also expandable to other devices if a stable network connection is available (Figure 1 and 2) Licensed under a Creative Commons Attribution 4.0 International License (CC BY 4.0). Attribution: owner/author(s). Web Audio Conference WAC-2025, November 19–21, 2025, Paris, France. © 2025 Copyright held by the owner/author(s). 3. METHODOLOGICAL CONSIDERRATIONS IN ENSEMBLE SETTINGS Previous studies have confirmed the effectiveness of the MVP Support System in controlled, individual-use environments. For example, experiments have compared performers’ ability to maintain visual engagement with a conductor or metronome displayed on a screen when using a traditional tuner versus wearing MVP-enabled smart glasses. However, applying the system in actual ensemble environments introduces additional methodological challenges due to the presence of multiple sound sources. A primary concern is audio interference—the possibility that microphones will capture sounds from adjacent performers rather than the target player. This issue is particularly acute with instruments like trumpets and trombones, which project sound forward, increasing the risk of cross-talk between neighboring players. To address this, the current study focused on a French horn ensemble. The French horn’s unique acoustic design features a backward-facing bell positioned to the player’s rear-right, which reduces the likelihood of picking up other players' sounds. Furthermore, horn ensembles are typically arranged in a semicircle, making it feasible to position the microphone behind each player in alignment with their instrument's sound projection. For the experimental trial, a horn quartet from a university ensemble was invited to participate. Each of the four performers took turns using the MVP Support System while performing the piece. The results demonstrated that the system successfully captured the target performer’s sound and delivered real-time feedback without interference from the other players. Furthermore, all four participants reported that the system was beneficial to their performance and operated effectively throughout the trial. 4. DISCUSSION AND FUTURE DIRECTIONS The findings suggest that the MVP Support System, when used in conjunction with smart glasses, can function effectively even in real ensemble rehearsal environments. User feedback aligned with results from prior laboratory-based studies, confirming the system’s usability and effectiveness. Future enhancements may include adjustable pitch tolerance modes—for example, a wider tolerance range for beginners and a stricter range for professional-level performance. Expanding device compatibility and improving noise-canceling algorithms will also be essential for enabling broader adoption, especially in larger ensembles or noisy environments. Additionally, given the physical limitations associated with microphone placement, future development should explore options such as instrument-mounted clip-on microphones, wireless transmission between microphone and computer, or improvements to the built-in microphones in smart glasses themselves. These developments will be critical for ensuring accurate, individualized pitch feedback without compromising performance comfort or ensemble coherence. 5. ACKNOWLEDGMENTS This research represents an exploratory effort to further develop and expand upon the author’s previous research topics [2][7]. This research is an extension of ongoing studies on the MVP Support System conducted by the author. It has been supported by JSPS KAKENHI Grant Number JP24K16761 6. REFERENCES [1] S. J. Morrison and J. Fyk. 2002. Intonation. In The Science and Psychology of Music Performance: Creative Strategies for Teaching and Learning, G. McPherson and R. Parncutt (Eds.). Oxford University Press, New York, NY, 183–198. [2] Y. Yamaguchi et al. 2022. A pilot study of the MVP support system using Google Glass. In Proceedings of EdMedia + Figure 1. Schematic of MVP support system Figure 2. Screenshot of the system Innovate Learning. Association for the Advancement of Computing in Education (AACE), 17–28. [3] Y. Yamaguchi. 2023. A brief scoping review of musical performance support system in IEEE study fields. In Proceedings of the 2023 IEEE International Conference on Teaching, Assessment and Learning for Engineering (TALE). [4] ml5.js. 2022. Pitch Detection API. NYU ITP. Retrieved from https://ml5js.org/ [5] J. W. Kim, J. Salamon, P. Li, and J. P. Bello. 2018. CREPE: A convolutional representation for pitch estimation. In Proceedings of the IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP). IEEE. [6] C. H. Chan. 2022. Simple Score Follower: A contextual switching approach to polyphonic score following on the web using deep-learning pitch detection. PhD Dissertation. Harvard University. [7] Y. Yamaguchi and T. Kawada. 2024. Enhancing wind instrumentalists’ performance: Evaluating the MVP support system with OpenPose motion capture. In Proceedings of the International Conference on Genetic and Evolutionary Computing. Springer Nature Singapore, Singapore, 352–359.