Echo residue judgment method

Abstract

The invention provides an echo residual judgment method in the field of echo cancellation. The echo residue judgment method comprises the steps of S10, confirming a far-end signal corresponding to a near-end signal and an error signal; S20, performing FFT on the three signals to obtain frequency domain signals, and further calculating the power spectral density and the cross-power spectral densityof each frequency point of the three signals; S30, calculating coherence coefficients of the near-end signal and the error signal as well as the far-end signal and the near-end signal; S40, dividingthe frequency domain part of the sound signal into three sub-bands, and calculating the average power spectral density of the error signal of each sub-band; S50, calculating an average coherence coefficient of each frequency band based on the coherence coefficient; S60, calculating the frequency spectrum flatness of the error signal of each frequency band based on the frequency domain signal, andsetting a threshold value; and S70, judging whether the error signal has echo residues or not based on the frequency spectrum flatness, the threshold value, the average power spectral density and theaverage coherence coefficient. The method has the advantages that the false detection rate of the echo state is greatly reduced, and then echo residues are reduced.

Description

technical field [0001] The invention relates to the technical field of echo cancellation, in particular to a method for judging residual echo. Background technique [0002] When the sound round-trip delay exceeds 30ms, the speaker can distinguish the echo from the local voice. The echo will degrade the sound quality of the call and affect the call experience. Therefore, the echo cancellation function is a necessary audio processing function in call applications. Video conferencing, loudspeaker and cell phone calls, etc. By eliminating the echo component in the signal picked up by the microphone, the interference of the echo to the call can be eliminated, the intelligibility of the voice can be improved, and howling can be prevented. [0003] Traditionally, the method of eliminating residual echo is based on the correlation between the error signal, the near-end signal and the far-end signal corresponding to the current near-end signal to generate a suppression factor, and a...

AI Technical Summary

Problems solved by technology

However, the traditional method has the following disadvantages: the judgment frequency band used is relatively single, non-speech frequency bands are omitted, or the coherence of all current frequency points is averaged, and the individual frequency points with high coherence become small and low in the overall situation. If it is lower than the threshold, it will lead to errors in the judgment of the echo state and the actual existence, miscalculation of the suppression curve of the error signal, resulting in residual echo, and even the howling caused by the self-excitation of the echo

Images