Title:
SOUND SIGNAL PROCESSING METHOD, SOUND SIGNAL PROCESSING APPARATUS AND SPEECH RECOGNIZER
Document Type and Number:
Japanese Patent JP3940662
Kind Code:
B2
Abstract:
PROBLEM TO BE SOLVED: To obtain a speech signal suitable as an input signal of speech recognition or the like by avoiding the problem of aliasing which occurs when removing noise from the speech signal mixed with the noise by using a microphone array.
SOLUTION: Sound signals are collected by a plurality of microphones, a target signal contained in the sound signals is emphasized and a sound signal suppressing a noise signal is obtained. Further, an output signal in which a specific frequency component is removed based on the arriving direction of the noise is produced from such a sound signal. Thus, the sound signal suitable as the speech signal to be used for the speech recognition or the like can be obtained.
Inventors:
Emperor Amada
Takataka Yamamoto
Takataka Yamamoto
Application Number:
JP2002333118A
Publication Date:
July 04, 2007
Filing Date:
November 18, 2002
Export Citation:
Assignee:
Toshiba Corporation
International Classes:
G10L15/20; G10L15/28; G10L21/02; G10L21/0208; G10L21/0232; H04R1/40; H04R3/04; (IPC1-7): G10L21/02; G10L15/20; G10L15/28; H04R1/40; H04R3/04
Domestic Patent References:
| JP10207490A | ||||
| JP7115316A | ||||
| JP2000354290A | ||||
| JP2002221988A |
Other References:
永田仁史 他,"話者追尾2チャネルマイクロホンアレーに関する検討",電子情報通信学会論文誌A,1999年 6月25日,Vol.J82-A,No.6,p.860-866
Attorney, Agent or Firm:
Hiroshi Horiguchi
Previous Patent: SWASH PLATE VOLUME VARIABLE CHAMBER TYPE FLUID MACHINERY
Next Patent: SIMULATION METHOD AND PROGRAM USING HYBRID MODEL
Next Patent: SIMULATION METHOD AND PROGRAM USING HYBRID MODEL