Processing Techniques for Intelligibility Improvement to Speech with Co-Channel Interference.

Abstract : Enhancing speech that has been corrupted during transmission or recording is a widely researched problem. A special case is when the interference is a second talker's voice. This report summarizes research on processing techniques that improve the intelligibility of the desired speech signal in the presence of such voice interference. Formal subjective intelligibility test procedures for evaluating enhancement algorithms for the voice interference problem are first developed. Computational objective measures based on spectral criteria are also developed to provide testing during intermediate stages of algorithm development.