Title:
【発明の名称】音声認識装置
Document Type and Number:
Japanese Patent JP2888556
Kind Code:
B2
Abstract:
PURPOSE:To prevent noise signal mixing in an audio signal converted to an electrical signal by inputting the voice of a recognized system in the mode of mechanical oscillation of a voice generation source. CONSTITUTION:When a voice recognition targeted person emits voice, the mechanical oscillation transmitted to a head part is detected with an oscillation sensor 6 adhered closely to the head part, and a discrete audio signal is outputted from a discrete value conversion part 7. A CPU 1 performs distance calculation between each reference pattern of the reference patterns stored in a ROM 2 and the feature pattern of the audio signal, and the voice emitted from the recognized system can be recognized by retrieving and extracting the character information of the reference pattern that entirely coincides with the feature pattern of the audio signal and executing such control procedure at a constant cycle repeatedly. By using such constitution, it is possible to make an ambient noise hard to be mixed, therefore, voice recognition ratio can be heightened.
Inventors:
TSUNODA MASAMI
Application Number:
JP25784389A
Publication Date:
May 10, 1999
Filing Date:
October 04, 1989
Export Citation:
Assignee:
KYANON KK
International Classes:
G10L15/28; (IPC1-7): G10L3/00; G10L3/00
Domestic Patent References:
| JP284906A | ||||
| JP5665200A | ||||
| JP6318400A | ||||
| JP59130141U |
Attorney, Agent or Firm:
Yoshikazu Tani