Bow Gesture Classification to Identify Three Different Expertise Levels: A Machine Learning Approach

Abstract

To acquire new skills in a high-level music context, students need many years of conscious dedication and practice. It is understood that precise motor actions have to be incorporated into the musicians’ automatic executions, where a repertoire of technical actions must be learned and mastered. In this study, we develop a computer modelled assistant applying machine learning algorithms, for self-practice musicians with the violin as a test case. We recorded synchronized data from the performer’s forearms implementing an IMU device with ambient sound recordings. The musicians perform seven standard bow gesture. We tested the model with three different expertise levels to identify relevant dissimilitudes among students and teachers.