Cooperative Automatic Accompaniment System Using Predictive Models of Expression in Music Performance Based on CRFs

Bibliographic Information

Other Title
  • CRFに基づく伴奏の演奏表現の予測モデルと協調演奏システム

Search this article


本稿では,複数パートを含む楽譜が与えられた際に,演奏者の一部演奏に合わせて,伴奏に適切な演奏表情を付与するための伴奏の予測モデルを提案する.複数パートを含む演奏の場合,それぞれのパートは旋律としての自然さを保ちながら,パート同士が調和して進行すると考える.本研究では,それらの関係を CRF(Conditional Random Fields,条件付き確率場) を用いて統計的に学習し,伴奏の演奏表現の予測モデルの予測精度について評価実験及び考察を行った.また,その応用として実時間で演奏位置を推定し,演奏表情が付与された伴奏を自動再生する協調演奏システムを提案する.In this paper, we propose the method to predict expression of accompaniment given a part of performance referred to the score which contains several parts. In corroborative performance, a part of performance has musical harmony as a melody and it harmonizes with other parts. In our approach, the harmonic relation between parts is modeled by CRF (Conditional Random Fields). Experimental results and evaluations of the accuracy of the predictive models which learned from data statistically are reported. Also, we present the cooperative automatic accompaniment system which estimates the performer's beat position in the score on real-time processing and play the expressive accompaniment automatically.



  • CRID
  • NII Article ID
  • NII Book ID
  • Text Lang
  • Data Source
    • CiNii Articles

Report a problem

Back to top