Title:
SPEECH RECOGNITION METHOD, DEVICE, AND APPARATUS, AND COMPUTER READABLE STORAGE MEDIUM
Document Type and Number:
WIPO Patent Application WO/2019/080248
Kind Code:
A1
Abstract:
A speech recognition method, device, and apparatus, and a computer readable storage medium. The speech recognition method comprises: performing, according to a preset rule, framing and acoustic feature extraction on speech information to be detected, to obtain a frame-level speech feature sequence (S101); sequentially dividing the frame-level speech feature sequence into n information blocks (S102); inputting all the information blocks in parallel to a preset bidirectional long short-term memory recurrent neural network model; and obtaining the output results of a forward recognition result and a backward recognition result corresponding to each information block under a corresponding neuron node of an output layer of the preset bidirectional long short-term memory recurrent neural network model, so as to obtain the speech recognition result of the speech information to be detected. The method can improve speech recognition effect, and shorten the time delay of speech decoding.
Inventors:
LIANG HAO (CN)
CHENG NING (CN)
WANG JIANZONG (CN)
XIAO JING (CN)
CHENG NING (CN)
WANG JIANZONG (CN)
XIAO JING (CN)
Application Number:
PCT/CN2017/113230
Publication Date:
May 02, 2019
Filing Date:
November 28, 2017
Export Citation:
Assignee:
PING AN TECH SHENZHEN CO LTD (CN)
International Classes:
G10L15/16; G10L15/26
Foreign References:
| CN105679316A | 2016-06-15 | |||
| CN104952448A | 2015-09-30 | |||
| CN105869624A | 2016-08-17 | |||
| US20120166194A1 | 2012-06-28 | |||
| CN106803422A | 2017-06-06 |
Attorney, Agent or Firm:
SHENZHEN TALENT PATENT SERVICE (CN)
Download PDF: