1996 Fiscal Year Final Research Report Summary
Sound Perception and Perceptual Integration Information
| Project/Area Number |
07308026
|
|---|
| Research Category |
Grant-in-Aid for Scientific Research (A)
|
| Allocation Type | Single-year Grants |
|---|
| Section | 総合 |
|---|
| Research Field |
Intelligent informatics
|
|---|
| Research Institution | Nagoya University |
|---|
Principal Investigator |
KAKEHI Kazuhiko Nagoya University, Graduate School of Human Informatics Professor, 大学院・人間情報学研究科, 教授 (90262930)
|
|---|
| Co-Investigator(Kenkyū-buntansha) |
KAWAHARA Hideki ATR Human Information Processing Research Laboratories, Department 1 Head, 人間情報通信研究所・第1研究室, 主幹研究員
AKAGI Masato Japan Advanced Institute of Science and Technology, Hokuriku, School of Informat, 情報科学研究科, 助教授 (20242571)
SHIMADA Shoji Ngaoka University of Technology, Faculty-of Engineering Professor, 工学部, 教授 (20262462)
SHIKANO Kiyohiro Advanced Institute of Science and Technology, Nara, School of Information Scienc, 情報科学研究科, 教授 (00263426)
ITAKURA Fumitada Nagoya University, School of Engineering Professor, 工学部, 教授 (30168299)
|
|---|
| Project Period (FY) |
1995 – 1996
|
| Keywords | Sound field / Sound loccalization / Reverberation suppression / Noise suppression / Sound source separation / Representation of sound signal / Synthesis of sound field / Perceptual integration of information |
|---|
| Research Abstract |
A fudamental research was conducted to clarify the robust human sound perception and cognition of sound environment, and to realize a new sound processing technique and information procesiing models for sound source localization, noise and reverberation supression, and synthesis of sound fields. Several new methods for sound localization and reverberation/noise suppression were developed. A new speech recognition system by HMM composition taking into the the some methods mentioned above was developed to prove its advantage to the usual system using reverberation/noise suppression technique merely as a pre-processor.
Individual study results are shwon as follows.
Sound field : ^<**>A design method of inverse filter to re-synthesize an original sound field in frequency domain. ^<**>A coding method for a sound field to be applied three dimensional space. ^<**>Data base of sound fields used for sound field simulation.
Sound localization : ^<**>An engineering model of a sound losalization using zero-crossing time information. ^<**>A good algorithm to calculate an inverse filter used for sound localization with a pair of head phone. ^<**>Finding a human ability of matching between mulitiple images of spekers and a speech sound source. ^<**>Finding that the perceptual distance to sound source depends on short time component of impulse response for imaginary sound sources.
Sound source separation : ^<**>The way of application of sound separation method of mutual information minimalization to actual sound fields. ^<**>Practical methods to suppress noise and reverberation using only 2 or 3 microphones. ^<**>A new extraction method of signal from the noise-added signal using human perceptual chracteristic.
Representation of sound signal : ^<**>A new high quality speech analysis-synthesis method by a new spectral envelope representation (STRAIGHT) and a new pitch extraction method (TEMPO)
|
