多重奏中の歌声の基本周波数と有声音素の同時推定手法

IPSJ

Bibliographic Information

Other Title
  • A Method for Concurrently Estimating F0 and Vowel Phoneme of Singing Voice in Polyphonic Music

Search this article

Description

本論文では，歌声の基本周波数（F0）と母音音素を同時に推定可能な新たな手法について述べる．本手法は，F0と母音音素だけでなく，歌手名や性別などの要素も同時に推定できるように設計されているため，混合音中の歌声を認識するための新たなフレームワークと考えることができる．本手法は，歌声とその他の伴奏音が混ざった状態を，歌声を分離するのではなく，そのままの形で統計的にモデル化する．また，信頼性の高い歌声のスペクトル包絡を推定するために，様々なF0を持つ複数の音の調波構造を使用する．F0と母音音素の同時推定を，ポピュラー音楽6歌手10曲で評価した結果，提案法によりF0推定の性能が平均3.7ポイント，音素推定の性能が平均6.2ポイント向上することを確認した．

A novel method is described that can be used to concurrently recognize the fundamental frequency (F0) and vowel phoneme of a singing voice (vocal) in polyphonic music. This method can be considered as a new framework for recognizing a singing voice in polyphonic music because it is designed to concurrently recognize other elements of a singing voice including singer's name and gender, though this paper focuses on the F0 and vowel phoneme. Our method stochastically models a mixture of a singing voice and other instrumental sounds without segregating the singing voice. It can also estimate a reliable spectral envelope by estimating it from the harmonic structure of many voices with various F0s. The experimental results of F0 and phoneme recognition with 10 popular-music songs by 6 singers showed that our method improves the recognition accuracy by 3.7 points for F0 estimation and 6.2 points for the phoneme recognition.

Journal

Keywords

Details 詳細情報について

Report a problem

Back to top