Acoustic Array Systems

7

Time-Domain MVDR Array Filter for Speech Enhancement

An excellent application of microphone arrays is speech enhancement, which consists of noise reduction and speech dereverberation. But sometimes in the literature, speech enhancement refers to noise reduction only.

Sensor arrays are already deployed in many real-world applications, where there is the need to extract a desired speech signal from competitive interferences and noises. However, the desired signal is never picked up alone by the microphones. Indeed, it is well known that the observations are usually very noisy and reverberant. Therefore, some intelligent processing has to be performed on all sensors in such a way that noise and reverberation are reduced without much distorting the desired signal. The conventional approach to dealing with these problems is by linear filtering. In this chapter, we study the powerful time-domain multichannel MVDR array filter for speech enhancement and show how it is related to some other optimal array filters. We also discuss many important recent developments and concepts, making this part an overview of the time-domain approaches for speech enhancement.

7.1 Signal Model and Problem Formulation

We consider the classical signal model in which a microphone array with sensors captures a convolved source signal in some noise field. The received signals at the discrete-time index are ...