]>

Figure 2: Diagram of the proposed encoding approach. Each of the microphone signals of the multimicrophone recording is processed by an analysis filterbank. Each of the subband signals is modeled using the source/filter model, and for each microphone signal, only the AR parameters are encoded. Recording 1 corresponds to the reference signal, which is encoded as a single audio channel; the residual of the reference signal will be used to resynthesize the remaining microphone signals at the receiving end.