Multichannel Signal Enhancement Using a Remote Wireless Microphone

Abstract:
A novel approach to multichannel signal enhancement is presented that exploits data from a remote wireless microphone (RWM). This RWM is placed near an interfering source and transmits only autocorrelation data of its observations to a host, i.e., not the entire signal. The host has access to the autocorrelation data from the RWM and measurements from a local microphone array. A blind source extraction algorithm exploits this correlation data in order to identify the desired extraction filter. This novel approach has several advantages compared to transmitting the entire signal of the RWM. By transmitting only correlation data the required transmission power reduces significantly and synchronization at sample level is not required. Furthermore, leakage of the desired source into the RWM does not degrade performance; and finally, both the desired and interfering sources may be active simultaneously. Index Terms — Blind source extraction, multichannel signal enhancement, remote wireless microphone, second order statistics