Login| Sign Up| Help| Contact|

Patent Searching and Data


Title:
TRAINING METHOD AND SPEECH RECOGNITION METHOD FOR HYBRID FREQUENCY ACOUSTIC RECOGNITION MODEL
Document Type and Number:
WIPO Patent Application WO/2018/153214
Kind Code:
A1
Abstract:
A training method and a speech recognition method for a hybrid frequency acoustic recognition model, which relate to the technical field of speech recognition. The method comprises: obtaining a first-type speech feature of a first speech signal and performing processing to obtain corresponding first speech training data (S1); obtaining a first-type speech feature of a second speech signal and performing processing to obtain corresponding second speech training data (S2); obtaining a second-type speech feature of the first speech signal and a second-type speech feature of the second speech signal according to power spectrums (S3); performing pre-training according to the first speech signal and the second speech signal, so as to form a preliminary recognition model of a hybrid frequency acoustic recognition model (S4); and performing supervised parameter training on the preliminary recognition model according to the first speech training data, the second speech training data and the second-type speech feature, so as to form a hybrid frequency acoustic recognition model (S5). The beneficial effects are as follows: the recognition model has good robustness and generalization.

More Like This:
WO/2006/059451SPEECH RECOGNITION DEVICE
JP2003099086METHOD, DEVICE AND PROGRAM FOR GENERATING LANGUAGE/ ACOUSTIC MODEL
JP2002311989SPEECH RECOGNITION METHOD WITH CORRECTED CHANNEL DISTORTION AND CORRECTED BACKGROUND NOISE
Inventors:
FAN LICHUN (CN)
Application Number:
PCT/CN2018/074320
Publication Date:
August 30, 2018
Filing Date:
January 26, 2018
Export Citation:
Click for automatic bibliography generation   Help
Assignee:
YUTOU TECH HANGZHOU CO LTD (CN)
International Classes:
G10L15/065
Foreign References:
CN105513590A2016-04-20
CN105590625A2016-05-18
CN101320560A2008-12-10
CN101014997A2007-08-08
US20090012785A12009-01-08
Other References:
HOKKING, RATTAPHON ET AL.: "Speech Recognition of Different Sampling Rates Using Fractal Code Descriptor", 2016 13TH INTERNATIONAL JOINT CONFERENCE ON COMPUTER SCIENCE AND SOFTWARE ENGINEERING (JCSSE, 13 July 2016 (2016-07-13) - 15 July 2016 (2016-07-15), XP033008971, Retrieved from the Internet
Attorney, Agent or Firm:
SHANGHAI SHENXIN LAW FIRM (CN)
Download PDF:
View/Download PDF      PDF Help



 
Previous Patent: MULTI-LANGUAGE HYBRID SPEECH RECOGNITION METHOD

Next Patent: METHOD FOR AUTOMATICALLY GENERATING SENTENCE SAMPLE WITH SIMILAR SEMANTICS