Multi-microphone acoustic echo cancellation using multi-channel warped linear prediction of common acoustical poles

A computationally cheap extension from single-microphone acoustic echo cancellation (AEC) to multi-microphone AEC is presented for the case of a single loudspeaker. It employs the idea of common-acoustical-pole and zero modeling of room transfer functions (RTFs). The RTF models used for multi-microphone AEC share a fixed common denominator polynomial, which is calculated off-line by means of a multi-channel warped linear prediction. By using the common denominator polynomial as a prefilter, only the numerator polynomial has to be estimated recursively for each microphone, hence adapting to changes in the RTFs. This approach allows to decrease the number of numerator coefficients by one order of magnitude for each microphone compared with all-zero modeling. In a first configuration, the prefiltering is done on the adaptive filter signal, hence achieving a pole-zero model of the RTF in the AEC. In a second configuration, the (inverse) prefiltering is done on the loudspeaker signal, hence achieving a dereverberation effect, in addition to AEC, on the microphone signals.