楽曲分類におけるラウド感を表す音響特徴量の評価  [in Japanese] Evaluation of the acoustic feature representing loud-ness of music signal for music classification  [in Japanese]

Search this Article

Author(s)

Abstract

本研究は,楽曲推薦の特徴量に利用出来るような音響特徴量,特にラウドロック系の楽曲を特徴付ける音響特徴量の提案をする.ラウドロック系の楽曲は,歪んだ音を出すエレキギター及びベース,激しく叩くドラムの音が特徴的であり,上記の楽曲と,それ以外の楽曲とを混在させた曲順で聴くことはほとんどない.よって,両者を分類できる特徴量は,有用な特徴量となることが期待できる.本稿では,提案した音響特徴量を用いて,ラウドロック系,J-pop,ニューエイジ系の自動ジャンル分類実験により客観評価を行った.最も良い条件の場合,ラウドロック系の楽曲とそれ以外の楽曲とを区別する2ジャンルの識別率が97%,3ジャンルの識別率が87%となった.さらに,この音響特徴量が楽曲の印象とどのような対応があるのか,印象尺度を用いた主観評価実験を行った.その結果,音響特徴量と「激しい」「騒々しい」「優しい」といった印象との間に,強い相関があることが確認され,提案する音響特徴量が,主観的尺度と対応することが明らかになった.

This paper proposes an acoustic feature measuring a degree of closeness to Loud-rock music for music genre classification. In music listening, many people do not use playlists including Loud-rock songs and other genre songs both, because Loud-rock songs contain really loud distortion guitars, electric basses and drums, quite differ in sound from other genre songs. Thus, an acoustic feature that can classify songs into the Loud-rock music and other genre music is useful. We performed an objective evaluation of the acoustic feature that classifies 104 songs into three classes of Loud-rock, Japanese-pop and New-age. As a result, the recognition rate of 97% was obtained in two class classification (Loud-rock and others), and that of 87% was obtained in three class classification. We also performed a subjective evaluation by using the impression scale measure in order to know subjective impression of the songs used in the objective evaluation. Since the results indicated strong correlation between the acoustic feature and three impression measures of Hageshii (fierce), Souzoushii (boisterous), and Yasashii (gentle), it was concluded that the proposed acoustic feature represents subjective impression appropriately.

Journal

  • IEICE technical report

    IEICE technical report 111(97), 41-46, 2011-06-16

    The Institute of Electronics, Information and Communication Engineers

References:  9

Codes

  • NII Article ID (NAID)
    110008746450
  • NII NACSIS-CAT ID (NCID)
    AN10013221
  • Text Lang
    JPN
  • Article Type
    ART
  • ISSN
    09135685
  • NDL Article ID
    11161301
  • NDL Source Classification
    ZN33(科学技術--電気工学・電気機械工業--電子工学・電気通信)
  • NDL Call No.
    Z16-940
  • Data Source
    CJP  NDL  NII-ELS 
Page Top