Speech recognition method, system, device, and storage medium

By using multiple microphones in the speech recognition system to collect sound source data, filter and fuse high-quality speech features, and combine preset network models for recognition, the problem of delay and quality differences in speech recognition under the multi-microphone array is solved, and the accuracy and efficiency of recognition are improved.

CN117935789BActive Publication Date: 2025-03-25CHINA UNICOM (GUANGDONG) IND INTERNET CO LTD
2 Cites 0 Cited by

Patent Information

Application Number
CN202410071729.1
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2024-01-17
Publication Date
2025-03-25
Estimated Expiration
2044-01-17

AI Technical Summary

Technical Problem

When existing speech recognition technology uses multi-microphone array to collect audio data, there are differences in delay and audio quality, which affects the recognition accuracy and processing efficiency.

Method used

The sound source data is collected through multiple microphones, voice features are extracted and preset features are compared, voice quality is analyzed, target sound source data that meets the quality requirements are selected, and their features are fused and processed, and voice recognition is performed in combination with the preset speech recognition network model.

Benefits of technology

The accuracy and processing efficiency of speech recognition are improved. By screening high-quality sound source data and fusion characteristics, the noise impact and data volume during the recognition process are reduced, and the overall recognition performance is improved.

✦ Generated by Eureka AI based on patent content.
Patent Text Reader

Abstract

The embodiment of the present application discloses a speech recognition method and system, device, and storage medium. The speech recognition method is applied to an electronic device included in a speech recognition system. The speech recognition system also includes a microphone array, and the microphone array includes multiple microphones. The method includes: obtaining multiple sound source data corresponding to each microphone through multiple microphones; obtaining the speech features corresponding to each sound source data, and comparing the similarity of the speech features corresponding to each sound source data with the preset speech features, and obtaining the matching speech features corresponding to each sound source data; performing speech quality analysis on the matching speech features corresponding to each sound source data, and determining the target sound source data according to the speech quality analysis results; performing fusion processing on the matching speech features corresponding to each target sound source data according to the speech quality analysis results corresponding to each target sound source data, and obtaining fusion speech features; obtaining speech recognition results according to the fusion speech features and the pre-set speech recognition network model. It is possible to perform speech quality analysis on the sound source data collected by different microphones, and perform fusion processing of speech features according to the results, and perform speech recognition by fusion speech features to improve the accuracy of speech recognition processing.
Need to check novelty before this filing date? Find Prior Art