摘要 |
<p>The method comprises, in the temporal domain, an echo cancellation processing (40) by subtracting from the signal picked up a reference signal delivered by an adaptive circuit for modelling the acoustic coupling between microphone (18) and loudspeaker (16). A processing for removing the residual echo (42) remaining after the echo cancellation, and a processing for reducing the noise (44) are operated in the frequency domain on successive frames of the signal. The residual echo removal processing comprises an estimation of the respective values of the power of the residual echo over a plurality frequency bands defined for successive frames of the signal, then the calculation and the application of a variable gain specific to each frequency band and to each frame, the respective gain being all the lower the higher the estimated relative power of the residual echo, for the frequency band considered, and vice versa. For each frequency band and for each frame, respective values of the power of the ambient noise present in the signal picked up are estimated a priori, and the relative levels of the power of the noise and of the power of the residual echo in the signal picked up are compared. The variable gain is dependent on the estimated relative power of the residual echo when the level of the latter is greater than that of the noise, and the estimated relative power of the noise in the converse case.</p> |