One-microphone enhancement of reverberant speech (2006)

Sound Demo for the Wu-Wang Reverberant Speech Enhancement System

The reverberant speech enhancement results for 8 sentences from 4 female and 4 male speakers using the Wu-Wang system described in the paper “A two-stage algorithm for one-microphone reverberant speech enhancement” by M. Wu and D.L. Wang (IEEE Trans. Audio, Speech, & Language Processing, vol. 14, pp. 774-784, 2006) are given in the following wave files.

The 1st and the 2nd columns (“Clean”) are clean speech utterances sampled at 16 kHz and 8 kHz, respectively. The 3rd and 4th columns (“REV”) are reverberant speech signals produced by convolving the clean signals and a room impulse response function with T60 = 0.3 s, sampled at 16 kHz and 8 kHz, respectively. The 5th column (“YM”) is the processed speech using the YM algorithm by B. Yegnanarayana and P.S. Murthy (described in “Enhancement of reverberant speech using LP residual signal,” IEEE Trans. Speech & Audio Processing, vol. 8, pp. 267-281, 2000) sampled at 8 kHz. Column 6 and 7 (“INV”) are inverse-filtered speech resulting from the first stage of the proposed algorithm sampled at 16 kHz and 8 kHz, respectively. Column 8 and 9 (“DEREV”) are the final processed speech using the proposed two-stage algorithm sampled at 16 kHz and 8 kHz, respectively.

Clean (16k) Clean (8k) REV (16k) REV (8k) YM (8k) INV (16k) INV (8K) DEREV (16K) DEREV (8K)
Female1 wav wav wav wav wav wav wav wav wav
Female2 wav wav wav wav wav wav wav wav wav
Female3 wav wav wav wav wav wav wav wav wav
Female4 wav wav wav wav wav wav wav wav wav
Male1 wav wav wav wav wav wav wav wav wav
Male2 wav wav wav wav wav wav wav wav wav
Male3 wav wav wav wav wav wav wav wav wav
Male4 wav wav wav wav wav wav wav wav wav