Title:
【発明の名称】音声切り出し装置
Document Type and Number:
Japanese Patent JP2648014
Kind Code:
B2
Abstract:
PURPOSE:To obtain the voice segmenting device which can set an appropriate threshold by taking a noise level picked up by a microphone for inputting a voice in consideration. CONSTITUTION:A parameter for representing a feature of a voice from a microphone 1 for converting a voice to an electric signal is extracted by a voice analyzing part 2, and switching of a switch S for selecting a sending-out destination of its feature parameter is controlled by a switching part 3. By a segmenting part 5, a processing for detecting a time area of the voice obtained through the switch S from the voice analyzing part 2 is executed, and it segmenting threshold is set and registered in a threshold memory 6. It is connected to the other switching terminal of the switch S, and the threshold to be registered in the memory is decided to an optimal value by a threshold learning part 7, based on a noise signal obtained from an acoustic input part 4 and a feature parameter obtained through the switch S. In such a way, a feature parameter time sequence segmented by the segmenting part 5 is used as a voice pattern required for a voice recognition processing by a pattern recognition.
More Like This:
Inventors:
TSURUFUJI SHINICHI
Application Number:
JP27839390A
Publication Date:
August 27, 1997
Filing Date:
October 16, 1990
Export Citation:
Assignee:
SANYO DENKI KK
International Classes:
G10L15/04; G10L25/78; (IPC1-7): G10L3/00
Attorney, Agent or Firm:
Koji Yasutomi (1 person outside)