System and process for regression-based residual acoustic echo suppression

A system for suppressing a portion of a microphone signal corresponding to a playback of the multiple channels of a speaker audio signal wherein each channel is played aloud via separate loudspeakers in a space having a microphone therein which picks-up said playback and generates a microphone signal, said system comprising:

• a general purpose computing device;
• a computer program comprising program modules executable by the computing device, said modules comprising,
• a multi-channel acoustic echo canceller (AEC) module which suppresses a first part of the playback of the multiple channels of the speaker signal found in the microphone signal and which generates an AEC output signal;
• a first regression-based residual echo suppression (RES) module which inputs the AEC output signal and one of the channels of the speaker signal, and suppresses at least a portion of a residual part of the speaker signal playback of the input channel found in the microphone signal which was left unsuppressed by the AEC module, to produce a RES output signal;
• and
• one or more additional RES modules each of which inputs the signal output by the preceding RES module and a channel of the speaker signal not yet input by a RES module, and suppresses at least a portion of a residual part of the speaker signal playback of the channel input thereby which was left unsuppressed by the AEC module and all preceding RES modules, to produce an additional RES output signal, and wherein the last signal output by the last RES module is designated as the final RES signal.

A system for suppressing a portion of a microphone signal corresponding to a playback of the right and left channels of a speaker stereo audio signal wherein each channel is played aloud via separate loudspeakers in a space having a microphone therein which picks-up said playback and generates the microphone signal, said system comprising:

• a general purpose computing device;
• a computer program comprising program modules executable by the computing device, said modules comprising,
• an stereo acoustic echo canceller (AEC) module which suppresses a first part of the playback of the left and right channels of the speaker signal found in the microphone signal and which generates an AEC output signal;
• a first regression-based residual echo suppression (RES) module which inputs the AEC output signal and one of the (i) sum of the right and left channels of the speaker signal or (ii) the difference of the right and left channels of the speaker signal, and suppresses at least a portion of a residual part of the speaker signal playback of the input channel sum or difference found in the microphone signal which was left unsuppressed by the AEC module, to produce a first RES output signal;
• and
• a second RES module which inputs the first RES output signal and either the sum or difference of the right and left channels of the speaker signal whichever was not input by the first RES module, and which suppresses at least a portion of a residual part of the speaker signal playback of the input channel sum or difference found in the microphone signal which was left unsuppressed by the AEC module and the first RES module, to produce a final RES output signal.