Piano Practice Evaluation and Visualization by HMM for Arbitrary Jumps and Mistakes
Matsuto Hori, Christoph M. Wilk, Shigeki Sagayama · 2019
In this paper, we present a piano practice assisting system that tracks the user's piano performance with mistakes (note insertions, deletions and substitutions) and arbitrary jumps (repeats and skips) using a HMM with a fast Viterbi decoding algorithm. After tracking the user's free practice, the system provides feedback on various aspects of his/her performance, including information on missed and mistakenly played notes, jumps in the performance, and note overlap for evaluation of legato and staccato articulation. Additionally, the system provides information for comparing the user's practice performance with a teacher's exemplary performance with regards to tempo and dynamics. The practice support system's usefulness was evaluated in a subjective experiment, in which piano students of varying experience used the system and gave feedback on its features.