Speech enhancement using linear prediction residual

Abstract

In this paper we propose a method for enhancement of speech in the presence of additive noise. The objective is to selectively enhance the high signal-to-noise ratio (SNR) regions in the noisy speech in the temporal and spectral domains, without causing signi®cant distortion in the resulting enhanced speech. This is proposed to be done at three di erent levels. (a) At the gross level, by identifying the regions of speech and noise in the temporal domain. (b) At the ®ner level, by identifying the regions of high and low SNR portions in the noisy speech. (c) At the short-time spectrum level, by enhancing the spectral peaks over spectral valleys. The basis for the proposed approach is to analyze linear prediction (LP) residual signal in short (1±2 ms) segments to determine whether a segment belongs to a noise region or speech region. In the speech regions the inverse spectral¯atness factor is signi®cantly higher than in the noisy regions. The LP residual signal enables us to deal with short segments of data due to uncorrelatedness of the samples. Processing of noisy speech for enhancement involves mostly weighting the LP residual signal samples. The weighted residual signal samples are used to excite the time-varying all-pole ®lter to produce enhanced speech. As the additive noise level in the speech signal is increased, the quality of the resulting enhanced speech decreases progressively due to loss of speech information in the low SNR, high noise regions. Thus the degradation in performance of enhancement is graceful as the overall SNR of the noisy speech is decreased. Ó

Combining cepstral normalization and cochlear implant-like speech processing for microphone array-based speech recognition

“…It might also be useful to compare the WER reduction gain, achieved by combining cepstral normalization and cochlear implant-like speech processing, with that achieved with the combination of cepstral normalization and other temporal domain processing, for instance the speech enhancement using linear prediction residual [22], etc. Moreover, combining cepstral normalization and cochlear implant-like speech processing might be beneficial in real situations where there is overlapping speech, e.g.…”

Recognition of consonant-vowel (CV) units under background noise using combined temporal and spectral preprocessing

“…Temporal processing Fig. 1 Block diagram of the combined temporal and spectral noisy speech enhancement method (Krishnamoorthy and Prasanna 2011) methods are based on identifying and enhancing the speechspecific regions of noisy speech (Yegnanarayana et al 1999Yegnanarayana and Murthy 2000). The merit of temporal processing is in the enhancement of speech specific regions and don't require explicitly modeling of degradation.…”

A New Sparse Representation Algorithm for Speech Denoising

“…At present, eliminating the background noise and improving the speech quality become a significant research direction. Recently, the speech denoising method such as the traditional spectral subtraction method [1][2], wavelet transform method [3][4], linear forecast analysis [5], the subspace denoising method [6], and various improvement denoising algorithm [7][8]. However, these methods exit a lot of defects.…”