Automatic Decision of Piano Fingering Based on a Hidden Markov Models

This paper proposes a Hidden Markov Model (HMM)-based algorithm for automatic decision of piano fingering. We represent the positions and forms of hands and fingers as HMM states and model the resulted sequence of performed notes as emissions associated with HMM transitions. Optimal fingering decision is thus formulated as Viterbi search to find the most likely sequence of state transitions. The proposed algorithm models the required efforts in pressing a key with a finger followed by another key with another finger, and in two-dimensional positioning of fingers on the piano keyboard with diatonic and chromatic keys. Fundamental functionality of the algorithm was verified through experiments with real piano pieces. This framework can be extended to polyphonic music containing chords.