ECHO SUPPRESSION USING ADAPTIVE GAIN BASED ON RESIDUAL ECHO ENERGY

摘要

An echo suppressor for bi-directional communications includes an echo canceller and a residual echo suppressor having variable gain. The gain of the residual echo suppressor is controlled based on a characteristic, such as the estimated energy, of the residual voice echo, or alternatively the entire residual echo, preferably on a per-frame basis. In some embodiments, the residual voice echo energy is compared against the estimated non-echo energy to determine the required amount of gain to apply so as to apply so as to attenuate the residual voice echo below a psychoacoustic perception level. In some optional embodiments, comfort noise is added to the output signal after the residual echo suppressor in an amount that corresponds to the amount of signal energy loss through the residual echo suppressor. Thus, in some embodiments, desired voice and background noise (including local background noise and comfort noise) are used to mask the presence of residual echo. This approach avoids the problematic differentiation between echo single-talk and double-talk (desired voice) situations, thereby obviating the need for a desired voice detector.