A method of generating one-hand piano score for disabled persons of the hand and fingers

2021 Fiscal Year Final Research Report

• Project/Area Number: 19K12248
• Research Category: Grant-in-Aid for Scientific Research (C)
• Allocation Type: Multi-year Fund
• Section: 一般
• Review Section: Basic Section 62030:Learning support system-related
• Research Institution: The University of Tokushima
• Principal Investigator: OHNO Masaki 徳島大学, 大学院社会産業理工学研究部(理工学域), 講師 (90433739)
• Co-Investigator(Kenkyū-buntansha): 獅々堀 正幹 徳島大学, 大学院社会産業理工学研究部(理工学域), 教授 (50274262)
• Project Period (FY): 2019-04-01 – 2022-03-31
• Keywords: 手指障害支援 / 演奏支援 / 音楽生成 / 楽譜生成 / 機械学習

Outline of Final Research Achievements

In the first year, the "HMM-based one-handed musical score generation model" was proposed, in which note sequences of two-handed musical scores were assigned to HMM observation symbol sequences and note sequences of one-handed musical scores were assigned to hidden state sequences, and melody and accompaniment were modeled separately. In the next year, we proposed a "Linked HMM" that takes into account the co-occurrence probability between the hidden state sequences of the melody generation model and the accompaniment generation model, and was able to reduce the generation of dissonance. In the final year, an "accompaniment selection model" was proposed to select an accompaniment series that maximizes the probability of generating the main melody series, thereby solving the problem of generating dissonant or unnatural rhythmic accompaniments.
In the evaluation experiment, reproducibility and audibility were obtained with an F-measure of 0.29 and 3.75, respectively.

Free Research Field

音楽情報科学

Academic Significance and Societal Importance of the Research Achievements

楽器を演奏することは、豊かな生活をおくるための手段の一つであるが、手指障害者が演奏しやすいように編曲された楽譜は少ないため、健常者との間に差が生じていた。本研究では、隠れマルコフモデルを用いて手指障害者向け楽譜の運指モデルを構築し、両手演奏楽譜を片手演奏楽譜へ自動変換する技術を考案した。本研究は、手指障害者の楽器演奏を支援するシステムとして、国内外で初めての取り組みであり、教育・医療・福祉分野に対する貢献が期待できる。