A novel a priori SNR estimation approach based on selective cepstro-temporal smoothing

Online Monaural Speech Enhancement Based on Periodicity Analysis and A Priori SNR Estimation

“…When applying this gain, only the periodic frames of the noisy speech are enhanced. Panel (c) shows the result by applying the gain calculated by (30). When applying this gain, both the periodic and the aperiodic frames of the noisy speech are enhanced.…”

“…where ( , , ) is the enhanced signal unit after applied the gain in (30). Please note that the maximal value of ̂0 ( ) (corresponding to 14.3 ms) is smaller than N/2 (corresponding to 16 ms); and this comb-filtering will not introduce signal delay in this frame-based processing.…”

“…One state-of-the-art statistical-model based minimum mean-square error (MMSE) monaural speech enhancement algorithm was also evaluated as a comparison. This MMSE algorithm includes a recently developed MMSE-based estimation algorithm of noise power [4] and a cepstro-temporal smoothing estimation algorithm of the a priori SNR [30]. The implementation of the algorithm is a MATLAB function provided by the author.…”

“…where 2 is a smoothing factor and empirically selected as 0.96, and is the final Wiener gain calculated as (30). On the right side of (27), the first term [ ( − 1, ) • ( − 1, )] can be seen as the estimated speech energy level in the previous frame without smoothing; the second term [ ( , ) −̂( , )] can be seen as the maximum likelihood estimation of the speech energy level in current frame, which will always be larger than zero as ̂ has been limited to in (26).…”

Determination of pitch range based on onset and offset analysis in modulation frequency domain

“…Pitch determination is a fundamental problem that attracts much attention in speech analysis. A robust pitch detection algorithm (PDA) is needed for many applications including computational auditory scene analysis (CASA), prosody analysis, speech enhancement/separation [2], speech recognition, and speaker identification [3].…”