An evaluation of stereo speech enhancement methods for different audio-visual scenarios

Following speech on TV or radio in the presence of interferers is sometimes challenging, in particular for the elderly and the hearing-impaired. To evaluate the performance of speech enhancement methods for such scenarios, we consider a stereo mixture composed of a speech signal and interfering sources. We apply different approaches to separate the mixture into two components, where the first component contains mainly speech (the desired signal) and the second component contains the rest of the mixture. An improved stereo signal is constructed by recombining these components such that speech gets emphasized with respect to the rest of the mixture and at the same time the amount of artifacts is kept to a minimum. Listening tests and objective measures show that the center extraction approach is in general the most effective, although it is sensitive to speaker positioning.