Noise cancellation techniques in speech signals (DLPNC)

A new noise cancellation technique “delayed linear prediction noise cancellation” (DLPNC) was described by the author (see ibid., p.WM9.25.1-25.4, 1992) in a prevoius paper. In this technique, the current clean speech sample is predicted as a linear combination of the past M noisy speech samples delayed by one pitch period T. That is why this technique is named “delayed linear prediction noise cancellation”. In this article, the DLPNC technique is compared with two well-known noise cancellation techniques, Wiener noise cancellation, formulated by Sambur (1979) and adaptive noise cancellation proposed by Widrow et. al. (1975) and modified by Sambur (1978). The quality of the processed speech was judged by informal perceptual listening tests throughout this work. The criteria being the quality of the processed speech and the computational load the preliminary results reported in this paper indicate that DLPNC ranks close to WNC and ANC in its noise cancellation capabilities