Performance analysis of multi-channel adaptive algorithms operating in a non-stationary acoustic teleconferencing setup

Several important issues in multi-channel signal processing are discussed. It is shown that in some situations nonstationarity may help to solve problems that are unsolvable when assuming a stationary setup. In addition, the behavior of several adaptive algorithms operating in such non-stationary environments is analyzed and explained. The ideas are illustrated using the acoustic teleconferencing setup as a framework. Similar principles can be applied to other problems in acoustic telecommunication. Since the main goal of this paper is to provide insight, most of the theory is developed in terms of Wiener filtering and ensemble averaging operations.