Speech recognition method, system and equipment

The invention provides a voice recognition method, system and equipment, and the method is applied to the technical field of voice recognition, and comprises the steps: receiving original audio data, and extracting voice features from the original audio data; encoding the voice features by using a s...

Full description

Saved in:
Bibliographic Details
Main Authors CHEN LINLI, NING YIJIAN, FENG JING, LUO ZHONGMING, LAN LIANG
Format Patent
LanguageChinese
English
Published 20.10.2023
Subjects
ACOUSTICS
MUSICAL INSTRUMENTS
PHYSICS
SPEECH ANALYSIS OR SYNTHESIS
SPEECH OR AUDIO CODING OR DECODING
SPEECH OR VOICE PROCESSING
SPEECH RECOGNITION
Online AccessGet full text

Cover

More Information
Summary:The invention provides a voice recognition method, system and equipment, and the method is applied to the technical field of voice recognition, and comprises the steps: receiving original audio data, and extracting voice features from the original audio data; encoding the voice features by using a shared encoder, and converting the voice features into vector features; according to the vector features, decoding by using a time sequence classification decoder to obtain a first decoding result; according to the vector features, decoding by using a self-attention decoder to obtain a second decoding result; the self-attention decoder and the time sequence classification decoder share parameters of the shared encoder; and generating a voice recognition result of the original audio data according to the first decoding result and the second decoding result. Through a cascade structure of the self-attention decoder and the time sequence classification decoder, decision fusion of decoding results is realized, so that a
Bibliography:Application Number: CN202310980820