An echo cancellation method and system based on neural network dual-talker detection

The RNN-based double-talk detection enhances AEC by dynamically adjusting filter coefficients and incorporating auditory masking effects, addressing the robustness and accuracy issues in existing AEC methods, resulting in improved echo suppression.

CN115457928BActive Publication Date: 2025-07-15CHENGDU SHUSHENG LANGLANG TECHNOLOGY CO LTD
1 Cites 0 Cited by

Patent Information

Application Number
CN202210888604.9
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2022-07-27
Publication Date
2025-07-15
Estimated Expiration
2042-07-27

AI Technical Summary

Technical Problem

Among the existing echo cancellation methods, the double-talk detection scheme has poor robustness and low detection accuracy, resulting in unsatisfactory echo cancellation effect, and the linear adaptive filter cannot effectively remove linear and nonlinear residual echoes.

Method used

A dual-talk detection method based on neural network is adopted, combined with recurrent neural network (RNN) for vocal detection, the dual-talk detection state results are controlled by a finite state machine, combined with linear adaptive filtering and nonlinear post-processing, and the filter coefficients are optimized using variable step factor and gain function to perform adaptive updates and residual echo suppression.

Benefits of technology

It improves the robustness and accuracy of double-talk detection, effectively removes linear and nonlinear residual echoes, improves the echo cancellation effect, prevents the adaptive filter from diverging when there is no human voice, and optimizes the fidelity of the near-end signal.

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

Abstract

The present invention provides an echo cancellation method and system based on neural network double-talk detection, which relates to the technical field of audio processing. The present invention controls whether to update the adaptive filter in the linear pre-processing by using the double-talk detection result; when it is detected that there is no vocal signal at the far end, the update of the adaptive filter is stopped. This prevents the adaptive filter from diverging due to perturbation and deviating from the stable point when in the double-talk state or when there is only ambient noise at the far end; by using the double-talk detection result to control the maximum suppression amount of the echo in the non-linear post-processing, when it is detected that the human ears at the near end and the far end are speaking simultaneously, the maximum suppression amount required to mask the residual echo is estimated in combination with the auditory masking effect of the human ear. While keeping the distortion of the useful signal at the near end small, the residual echo is also effectively suppressed; when it is detected that there is only a single-talk signal from the human ear at the far end, the suppression of the residual echo signal is enhanced, so that the residual echo can be completely removed.
Need to check novelty before this filing date? Find Prior Art