The problem of speech coding of sounds obtained from several microphones is adressed here. In view to improve the coding performances by accounting for the coupling that exists among the microphones, a. multichannel linear prediction approach is proposed. Encoding of the linear prediction coefficients from multichannel reflection coefficients is presented. In view to extend the concept of LSP representations to the vector case, some results about the multichannel line spectrum decompositions are mentionned. The relevance of the multichannel approach is discussed and checked on simulations and real signals.
Bibliographic reference. Chonavel, T. / Saoudi, S. (1995): "Multi-channel linear predictive coding of audio signals", In EUROSPEECH-1995, 53-56.