楽譜情報を用いない歌唱力自動評価手法  [in Japanese] An Automatic Singing Skill Evaluation Method for Unknown Melodies  [in Japanese]

Access this Article

Search this Article

Author(s)

    • 中野 倫靖 NAKANO TOMOYASU
    • 筑波大学大学院図書館情報メディア研究科 Graduate School of Library, Information and Media Studies, University of Tsukuba
    • 平賀 譲 HIRAGA YUZURU
    • 筑波大学大学院図書館情報メディア研究科 Graduate School of Library, Information and Media Studies, University of Tsukuba

Abstract

本論文では,歌唱力を自動的に評価するシステム開発の第1 段階として,ポピュラー音楽における歌唱力の「うまい」「へた」を,楽譜情報を用いずに自動的に識別する手法を提案する.従来,訓練された歌唱者の歌唱音声に関する音響学的な考察は行われてきたが,それらの研究は歌唱力の自動評価に直接適用されたり,人間による評価と結び付けて検討されたりすることはなかった.本論文では,聴取者の歌唱力評価の安定性を聴取実験によって確認し,そこで得られた結果から歌唱音声に「うまい」「へた」をラベル付けして自動識別実験を行った.そのための特徴量として,歌唱者や曲に依存しない特徴であることを条件に,相対音高とビブラートの2 つを提案する.聴取実験では,22 人の聴取者を被験者とし,聴取者間の評価に相関があった組の割合は88.9%(p < .05)であった.また,600 フレーズのラベル付けされた歌唱音声に対して識別実験を行った結果,83.5%の識別率を得た.As a first step towards developing an automatic singing skill evaluation system, this paper presents a method of classifying singing skills (good/poor) that does not require score information of the sung melody. Previous research on singing evaluation has focused on analyzing the characteristics of singing voice, but were not directly applied to automatic evaluation or studied in comparison with the evaluation by human subjects. In order to achieve our goal, two preliminary experiments, verifying whether the subjective judgments of human subjects are stable, and automatic evaluation of performance by a 2-class classification (good/poor ), were conducted. The approach presented in the classification experiment uses pitch interval accuracy and vibrato as acoustic features which are independent from specific characteristics of the singer or melody. In the subjective experiment with 22 subjects, 88.9% of the correlation between the subjects' evaluations were significant at the 5% level. In the classification experiment with 600 song sequences, our method achieved a classification rate of 83.5%.

As a first step towards developing an automatic singing skill evaluation system, this paper presents a method of classifying singing skills (good/poor) that does not require score information of the sung melody. Previous research on singing evaluation has focused on analyzing the characteristics of singing voice, but were not directly applied to automatic evaluation or studied in comparison with the evaluation by human subjects. In order to achieve our goal, two preliminary experiments, verifying whether the subjective judgments of human subjects are stable, and automatic evaluation of performance by a 2-class classification (good/poor), were conducted. The approach presented in the classification experiment uses pitch interval accuracy and vibrato as acoustic features which are independent from specific characteristics of the singer or melody. In the subjective experiment with 22 subjects, 88.9% of the correlation between the subjects' evaluations were significant at the 5% level. In the classification experiment with 600 song sequences, our method achieved a classification rate of 83.5%.

Journal

  • IPSJ journal

    IPSJ journal 48(1), 227-236, 2007-01-15

    Information Processing Society of Japan (IPSJ)

References:  25

Cited by:  14

Codes

  • NII Article ID (NAID)
    110006152199
  • NII NACSIS-CAT ID (NCID)
    AN00116647
  • Text Lang
    JPN
  • Article Type
    Journal Article
  • ISSN
    1882-7764
  • NDL Article ID
    8649894
  • NDL Source Classification
    ZM13(科学技術--科学技術一般--データ処理・計算機)
  • NDL Call No.
    Z14-741
  • Data Source
    CJP  CJPref  NDL  NII-ELS  IR  IPSJ 
Page Top