Title:
音声分析システムおよび音声分析方法
Document Type and Number:
Japanese Patent JP7002822
Kind Code:
B2
Abstract:
To provide "a speech analysis system and a speech analysis method" that enable a start point and an end point of a wake word to be detected with high precision.SOLUTION: A speech analysis system 1 comprises: a registered word detection unit 12 which compares a speech pattern of an input speech with those of registered specific words to detect a wake word; a frequency specification unit 15 which specifies, as a target frequency, a frequency corresponding to a frequency of the highest sound pressure level among frequencies constituting the input speech; and a point detection unit 16 which detects, based upon transition in sound pressure level of the target frequency in a period including the timing when a specific word is detected, the start point and end point of the specific word in the input speech, the respective points being detected targeting for analysis of a waveform of a user-characteristic frequency in a period in which the specific word may be included with a very high possibility.SELECTED DRAWING: Figure 1
Inventors:
Kudo Nobunori
Application Number:
JP2018228620A
Publication Date:
January 20, 2022
Filing Date:
December 06, 2018
Export Citation:
Assignee:
Alpine Electronics Co., Ltd.
International Classes:
G10L15/04; G10L15/10
Domestic Patent References:
| JP2004233541A |
Attorney, Agent or Firm:
Kazuyuki Tachibana
Kyozo Katayose
Katsushi Sato
Masahiko Amagai
Kyozo Katayose
Katsushi Sato
Masahiko Amagai
Previous Patent: Pressure sensor
Next Patent: Guidance voice output control system and guidance voice output control method
Next Patent: Guidance voice output control system and guidance voice output control method