This paper describes a new and realisable speech enhancement structure which simulates the cocktail party effect with a modified iterative Wiener filter and a multi-layer perceptron neural network. The key idea is to use the neural network as a speaker recognition system to govern the iterative Wiener filter. The neural network is a modified perceptron with a hidden layer using feature date extracted from LPC cepstral analysis. The proposed technique has been successfully used for speech enhancement when the interference is competing speech or broad band noise.

A microphone array system with multi-stage processing for speech enhancement is presented in this paper. Two beamformers with uniform directional patterns, one aimed at the target source and the other at the interfering sources, convert the multi-channel inputs into two data sequences. A novel microphone array structure with a small aperture has been designed to obtain the dual beamformers. The outputs of the two beam-formers are then presented to a post-processing stage to further improve the quality and intelligibility of the speech signal. The post-processing stage can be selected from one of three different algorithms that are presented, which are suitable for different acoustic environments. Applications for such a system include hands-free telephony, teleconferencing and also special situations where speech signals must be picked up in an extremely noisy acoustic environment in which the microphones are hidden (e.g. in a forensic covert recording system).