Robust acoustic echo cancellation for loosely paired devices based on semi-blind multichannel demixing

A method comprising:

• receiving a multichannel audio signal comprising a plurality of audio channels;
• receiving a reference signal;
• transforming each of the plurality of audio channels to a plurality of complex-valued subband audio signals, and transforming the reference signal to a corresponding plurality of complex-valued subband reference signals;
• producing a probability of acoustic echo dominance for each of the plurality of complex-valued subband audio signals based on the corresponding complex-valued subband reference signal;
• performing supervised multichannel linear processing controlled by the corresponding probability of acoustic echo dominance to decompose each complex-valued subband audio signal into a near-end source signal and a residual echo signal;
• postfiltering each complex-valued subband audio signal to remove residual echo signal components from the near-end source signal;
• and
• synthesizing the near-end source signals for each of the plurality of audio channels to produce an enhanced multichannel audio signal.