Misc - NISIMURA Ryuichi

TEMPORALLY VARIABLE MULTI-ASPECT AUDITORY MORPHING ENABLING EXTRAPOLATION WITHOUT OBJECTIVE AND PERCEPTUAL BREAKDOWN

H. Kawahara, R. Nisimura, T. Irino, M. Morise, T. Takahashi, H. Banno

2009 IEEE INTERNATIONAL CONFERENCE ON ACOUSTICS, SPEECH, AND SIGNAL PROCESSING, VOLS 1- 8, PROCEEDINGS ( IEEE )    3905 - +   2009  [Refereed]

　View Summary

A generalized framework of auditory morphing based on the speech analysis, modification and resynthesis system STRAIGHT is proposed that enables each morphing rate of representational aspects to be a function of time, including the temporal axis itself. Two types of algorithms were derived: an incremental algorithm for real-time manipulation of morphing rates and a batch processing algorithm for off-line post-production applications. By defining morphing in terms of the derivative of mapping functions in the logarithmic domain, breakdown of morphing resynthesis found in the previous formulation in the case of extrapolations was eliminated. A method to alleviate perceptual defects in extrapolation is also introduced.

Observation of empirical cumulative distribution of vowel spectral distances and its application to vowel based voice conversion

Hideki Kawahara, Masanori Morise, Toru Takahashi, Hideki Banno, Ryuichi Nisimura, Toshio Irino

INTERSPEECH 2009: 10TH ANNUAL CONFERENCE OF THE INTERNATIONAL SPEECH COMMUNICATION ASSOCIATION 2009, VOLS 1-5 ( ISCA-INT SPEECH COMMUNICATION ASSOC )    2615 - +   2009  [Refereed]

　View Summary

A simple and fast voice conversion method based only on vowel information is proposed. The proposed method relies on empirical distribution of perceptual spectral distances between representative examples of each vowel segment extracted using TANDEM-STRAIGHT spectral envelope estimation procedure [1]. Mapping functions of vowel spectra are designed to preserve vowel space structure defined by the observed empirical distribution while transforming position and orientation of the structure in an abstract vowel spectral space. By introducing physiological constraints in vocal tract shapes and vocal tract length normalization, difficulties in careful frequency alignment between vowel template spectra of the source and the target speakers can be alleviated without significant degradations in converted speech. The proposed method is a frame-based instantaneous method and is relevant for real-time processing. Applications of the proposed method in-cross language voice conversion are also discussed.

Development of Speech Input Method for Interactive VoiceWeb Systems

Ryuichi Nisimura, Jumpei Miyake, Hideki Kawahara, Toshio Irino

HUMAN-COMPUTER INTERACTION, PT II ( SPRINGER-VERLAG BERLIN )  5611   710 - +   2009  [Refereed]

　View Summary

We have developed a speech input method called "w3voice" to build practical and handy voice-enabled Web applications. It is constructed using a simple Java applet and CGI programs comprising free software. In Our website (http://w3voice.jp/), we have released automatic speech recognition and spoken dialogue applications that are Suitable for practical use. The mechanism of voice-based interaction is developed on the basis of raw audio signal transmissions via the POST method and the redirection response of HTTP. The system also aims at oroanizing a voice database collected from home and office environments over the Internet. The purpose of the work is to observe actual voice interactions of human-machine and human-human. We have succeeded in acquiring 8,412 inputs (47.9 inputs per day) captured by using normal PCs over a period of seven months. The experiments confirmed the user-friendliness of our system in human-machine dialogues with trial users.

Parameter optimization for a fundamental frequency extractor based on TANDEM-STRAIGHT

ITAGAKI Hanae, MORISE Masanori, NISHIMURA Ryuichi, IRINO Toshio, KAWAHARA Hideki

IPSJ SIG Notes ( Information Processing Society of Japan (IPSJ) )  2008 ( 123 ) 155 - 160   2008.12

　View Summary

A fundamental frequency extractor based on a temporally stable power spectral representation for periodic signals (TANDEM spectrum) and a spectral envelope derived from the representation (STRAIGHT spectrum) is proposed. This article describes roles of system parameters of the proposed method and their effects on system performance and reports results of preliminary optimization of them. System parameters investigated are; number of harmonic component for detecting hypothesized periodicity peak and weighting width on the log-lag domain for integrating specialized individual F0 detectors. Detailed descriptions of these parameters and their impact on F0 extraction performance are presented and they were further investigated using a database consisting of simultaneous recording of speech and EGG (electroglottogram) signals. Test results indicated that the proposed method has comparable performance with F0 extractors used in STRAIGHT and other popular F0 extractors such as YIN, when three harmonic components are used and weighting with a width of 1/√<2> of the center lag is used.

基本周波数情報に基づく線形予測と時間軸伸縮を利用した非周期成分の抽出について

河原 英紀, 森勢 将雅, 高橋 徹, 坂野 秀樹, 西村 竜一, 入野 俊夫

電子情報通信学会技術研究報告     Vol.108, No.338, SP2008-93/NLC2008-38   2008.12

劣化音声の知覚特性と音声認識器の認識傾向の比較

森本 隆司, 入野 俊夫, 西村 竜一, 河原 英紀

日本音響学会聴覚研究会資料 ( 日本音響学会聴覚研究委員会 )  38 ( 8 ) H-2008-142, Vol.38, No.8, pp.803-808 - 808   2008.12

TANDEM-STRAIGHTに基づく基本周波数抽出法に関する一検討

板垣 英恵, 森勢 将雅, 西村 竜一, 入野 俊夫, 河原 英紀

電子情報通信学会技術研究報告     Vol.108, No.338, SP2008-105/NLC2008-50   2008.12

実時間操作インタフェースへの応用を目的とした歌唱モーフィング操作パラメタの時系列への拡張について

河原英紀, 森勢将雅, 高橋徹, 坂野秀樹, 西村竜一, 入野俊夫

情報処理学会研究報告, 2008-MUS-78-16 ( Information Processing Society of Japan (IPSJ) )  2008 ( 127 ) 91-96 - 96   2008.12

　View Summary

Reuse of performance design in singing requires temporally localized manipulations of singing style, voice quality and expressions. They can be done in realtime such as in live concert scenes or can be done in off-line such as in the post production editing or recorded materials. A new framework is introduced to extend TANDEM-STRAIGTH-based morphing with a temporally variable multi-dimensional morphing rate and formulated. This formulation provides solid basis for implementing five morphing parameters (fundamental frequency, aperiodicity, STRAIGHT spectrogram, time and frequency axes) on each time-series independently. This formulation is based on interpolation of logarithmic derivative of transformation functions and enables extrapolative morphing without quality breakdown found in our previous formulations. The proposed method is easily extended to multiple exemplar morphing because the formulation is symmetric for each exemplar utterance.

パネル討論 「アプリケーション開発における音声認識」

西村 竜一, 中野 鐵兵, 栗原 一貴, 中臺 一博, 吉野 孝

情報処理学会研究報告     2008-SLP-73-11   2008.10

零周波数フィルタ信号に基づく基本周波数抽出法の評価と応用について

河原 英紀, 森勢 将雅, 高橋 徹, 坂野 秀樹, 大西 壮登, 板垣 英恵, 西村 竜一

日本音響学会2008年秋季研究発表会講演論文集     pp.423-424   2008.09

音声認識Webシステムにおける単語辞書構築技術

西村 竜一, 鈴田 健太郎, 河原 英紀, 入野 俊夫

日本音響学会2008年秋季研究発表会講演論文集     pp.197-198   2008.09

ウェブ上の言語知識を利用した音声認識用単語辞書の更新手法

鈴田 健太郎, 西村 竜一, 河原 英紀, 入野 俊夫

FIT2008 第7回情報科学技術フォーラム講演論文集(一般論文)     pp.189-190   2008.09

零周波数フィルタ信号に基づく基本周波数抽出法のTANDEM-STRAIGHTへの応用について

河原 英紀, 森勢 将雅, 坂野 秀樹, 高橋 徹, 板垣 英恵, 大西 壮登, 西村 竜一

情報処理学会研究報告     2008-MUS-76-17   2008.08

A temporally stable representation of power spectra of periodic signals and its application to F0 and periodicity estimation

Hideki Kawahara, Masanori Morise, Toru Takahashi, Ryuichi Nisimura, Hideki Banno, Toshio Irino

Acoustics'08 Paris     IpSCc24   2008.07  [Refereed]

A unified approach for F0 extraction and aperiodicity estimation based on a temporally stable power spectral representation

Hideki Kawahara, Masanori Morise, Toru Takahashi, Ryuichi Nisimura, Hideki Banno, Toshio Irino

ISCA Tutorial and Research Workshop (ITRW) on "Speech Analysis and Processing for Knowledge Discovery"     2008.06  [Refereed]

TANDEM-STRAIGHTによるスペクトル包絡の近似精度の改善について ～基本周波数により定まるNyquist周波数以上の空間周波数成分の復元について～

河原 英紀, 森勢 将雅, 高橋 徹, 坂野 秀樹, 西村 竜一, 入野 俊夫

電子情報通信学会技術研究報告   108 ( 116 ) SP2008-23, pp.19-24 - 24   2008.06

STRAIGHTを用いた簡易モーフィングによる印象変化の評価について

西田 沙織, 大西 壮登, 吉田 有里, 森勢 将雅, 西村 竜一, 入野 俊夫, 河原 英紀

情報処理学会研究報告     2008-HCI-128-8/2008-MUS-75-8, pp.43-48   2008.05

STRAIGHTに基づく柔軟な音声合成技術の開発

河原 英紀, 大西 壮登, 森勢 将雅, 高橋 徹, 西村 竜一, 坂野 秀樹, 入野 俊夫

情報処理学会第70回全国大会講演論文集     4L-5, vol.5, pp.357-358   2008.03

TANDEM およびSTRAIGHT スペクトルに基づく基本周波数および非周期性の表現について

河原 英紀,森勢 将雅,高橋 徹,西村 竜一,坂野 秀樹,入野 俊夫

日本音響学会2008年春季研究発表会講演論文集     pp.563-564 - 564   2008.03

音声入力Webシステムによる音声認識アプリケーションの構築技術

西村 竜一, 三宅 純平, 河原 英紀, 入野 俊夫

情報処理学会第70回全国大会講演論文集     3L-5, vol.5, pp.343-344   2008.03