A communication voice transmission method and system based on voice enhancement

By generating multi-turn interaction feature vectors through multi-channel speech acquisition and prediction networks, calculating the criticality and dynamic weight of future speeches, and adaptively adjusting enhancement parameters in real time, the dynamic adjustment problem of speech enhancement systems in multi-person, multi-turn interaction scenarios is solved, improving speech clarity and stability, and enhancing communication quality.

CN122245330APending Publication Date: 2026-06-19SHENZHEN YITENGJIE INFORMATION TECHNOLOGY CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
SHENZHEN YITENGJIE INFORMATION TECHNOLOGY CO LTD
Filing Date
2026-04-10
Publication Date
2026-06-19

AI Technical Summary

Technical Problem

Existing voice enhancement systems struggle to dynamically adjust enhancement strategies in multi-person, multi-turn interactive scenarios, resulting in the suppression of key statements or the weakening of background information, which affects the naturalness of the call and the integrity of the information.

Method used

By combining multi-channel speech acquisition with historical interaction information and network status, multi-round interaction feature vectors are generated. A prediction network is used to calculate the keyness and dynamic weight of future speeches. Real-time sequence modeling and adaptive adjustment of enhancement parameters are performed to build an adaptive judgment mechanism to improve speech clarity and stability.

Benefits of technology

In multi-person, multi-turn interaction scenarios, the targeting and stability of voice enhancement have been improved, the distortion and auditory discomfort caused by frequent fluctuations in enhancement parameters have been reduced, and the communication quality has been improved.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN122245330A_ABST
    Figure CN122245330A_ABST
Patent Text Reader

Abstract

This invention relates to the field of voice transmission in communication and discloses a voice transmission method and system based on voice enhancement. The method includes: real-time acquisition of multi-channel voice streams via a communication terminal; extraction of spectral, temporal, and acoustic features from the voice signal by combining historical multi-round interaction information and network status; generation of multi-round interaction feature vectors; calculation of the criticality of future rounds of speech using a prediction network; outputting prediction results and weights; inputting the prediction results and weights into an end-to-end voice enhancement network to generate an enhancement parameter set; real-time analysis of the change trajectory of the enhanced voice signal and the corresponding enhancement parameter set within a continuous time window; conversion of the analysis results into updated prediction weights and enhancement parameters using a mapping function; generation of a corrected decision curve based on the updated weight parameters and enhancement parameters; and outputting the updated decision curve and dynamic mapping data. This invention has the advantage of improving the relevance and stability of voice enhancement in multi-person communication scenarios.
Need to check novelty before this filing date? Find Prior Art