VocaListener:ユーザ歌唱を真似る歌声合成パラメータを自動推定するシステムの提案  [in Japanese] VocaListener : An Automatic Parameter Estimation System for Singing Synthesis by Mimicking User's Singing  [in Japanese]

Access this Article

Search this Article

Author(s)

Abstract

本研究では、歌声合成を使用した音楽制作を支援するために、ユーザの歌唱音声から歌声合成パラメータを自動推定するシステム VocaListener を提案する。従来、ユーザの歌唱音声から、音高や音量等を推定して歌声合成パラメータとする研究はあったが、歌声合成の条件 (歌声合成システムやその音源データ) の違いに対してロバストでなく、入力歌唱を真似るだけでは、ユーザの歌唱力を超えることが出来ないという問題もあった。そこで VocaListener では、合成された歌唱が入力歌唱と近くなるように、合成パラメータを反復更新することで、上記の条件の変化へ対処する。さらに、入力歌唱に対して、音高のずれやビブラートなどの歌唱要素を修正できる支援機能も提供する。本稿では、市販の歌声合成システムを対象に、音高と音量に関する合成パラメータを推定した結果を報告する。To support use of singing synthesis in producing music, we propose a system called VocaListener that automatically estimates parameters for singing synthesis from user's singing voice. Although there was a previous method that can estimate singing synthesis parameters of the F0 and power from user's singing voice, it was not robust with respect to different singing synthesis conditions (e.g., singing synthesis systems and their singer databases) and had difficulty overcoming limitations of user's singing skill. To deal with those different conditions, VocaListener repeatedly updates singing synthesis parameters so that synthesized singing can be closer to the user's singing. Moreover, we provide functions that help to modify the user's singing by correcting off-pitch phrases or changing vibrato. By using a singing synthesis system on the market, this paper reports results of estimating synthesis parameters of the F0 and power.

To support use of singing synthesis in producing music, we propose a system called VocaListener that automatically estimates parameters for singing synthesis from user's singing voice. Although there was a previous method that can estimate singing synthesis parameters of the F0 and power from user's singing voice, it was not robust with respect to different singing synthesis conditions (e.g., singing synthesis systems and their singer databases) and had difficulty overcoming limitations of user's singing skill. To deal with those different conditions, VocaListener repeatedly updates singing synthesis parameters so that synthesized singing can be closer to the user's singing. Moreover, we provide functions that help to modify the user's singing by correcting off-pitch phrases or changing vibrato. By using a singing synthesis system on the market, this paper reports results of estimating synthesis parameters of the F0 and power.

Journal

  • IPSJ SIG Notes

    IPSJ SIG Notes 2008(50(2008-MUS-075)), 49-56, 2008-05-21

    Information Processing Society of Japan (IPSJ)

Cited by:  7

Codes

  • NII Article ID (NAID)
    110006794542
  • NII NACSIS-CAT ID (NCID)
    AN10438388
  • Text Lang
    JPN
  • Article Type
    Journal Article
  • Data Source
    CJPref  NII-ELS  IPSJ 
Page Top