楽譜を考慮した演奏者の統計的モデル化手法の改善と演奏者識別による評価  [in Japanese] Improvement in stochastic performer modeling reflects musical score, and its evaluation by performer identification  [in Japanese]

Search this Article

Author(s)

Abstract

個性を備えた演奏は,楽譜の演奏指示に対して演奏者の意図による逸脱が加わることで生じると考えられる.我々はそのような逸脱の振る舞いが確率的な偏りを持つと仮定し,実演奏とその楽譜に基づいて両者の依存関係を統計学習する手法を提案してきた.従来法では楽譜から扱える演奏指示の種類が限られており,演奏指示の限られる楽曲には柔軟な対応ができないことが課題となっていたが,本稿では音符単位で付与できる演奏指示情報を拡充し,より楽曲に適合したモデルの生成について検討した.また,演奏者モデルの木構造についてもより効率的な生成手法を検討し,得られた演奏者モデルに未知の演奏における演奏者を識別させる評価を行った.これによって提案手法の有効性を検討し,今後の課題についても述べる.A unique performance is caused because deviating by the performer's intention joins the instruction of the score. We assumed that have bias with a stochastic behavior of such deviating. And we have already proposed a technique to train dependence of both as stochastic models from performance and its score. In a past proposal, performance instructions dealed from the score were not enough, and being not able to do flexible correspondence became a problem in music from which the performance instruction was limited. In this paper, the generation of the model with high adaptability to music was examined. Moreover, a more efficient generation technique was examined about the tree structure of the performer model, and the evaluation to identify the performer in an unseen performance was done to the trained performer model. The effectiveness of the proposal technique is examined, and future tasks are described from results.

Journal

  • 情報処理学会研究報告. [音楽情報科学]

    情報処理学会研究報告. [音楽情報科学] 86, Y1-Y7, 2010-07-28

    情報処理学会

References:  10

Codes

  • NII Article ID (NAID)
    110007997438
  • NII NACSIS-CAT ID (NCID)
    AN10438388
  • Text Lang
    JPN
  • Article Type
    ART
  • ISSN
    09196072
  • NDL Article ID
    025080603
  • NDL Call No.
    YH247-911
  • Data Source
    CJP  NDL  NII-ELS 
Page Top