Subs-Espec

ENHANCEMENT OF SPEECH CORRUPTED BY ACOUSTIC NOISE* M. Berouti, R. Schwartz, and J. Makhoul Bolt Beranek and Newman Inc. Cambridge, Mass.

ABSTRACT This paper describes a method for enhancing speech corrupted by broadband noise. The method is based on the spectral noise subtraction method. The original method entails subtracting an estimate of the noise power spectrum from the speech powerspectrum, setting negative differences to zero, the recombining the new power spectrum with original phase, and then reconstructing the time waveform. While this method reduces the broadband noise, it also usually introduces an annoying "musical noise". We have devised a method that eliminates this "musical noise" while further reducing the background noise. The method consists in subtracting anoverestimate of the noise power and preventing the resultant spectral spectrum, components from going below a preset minimum level (søectral floor). The method can automatically adapt to a wide range of signal—to—noise ratios, as long as a reasonable estimate of the noise spectrum can be obtained. Extensive listening tests were to determine the and performed quality intelligibility of speech enhanced byour method. Listeners unanimously preferred the quality of the for an processed speech. Also, input signal—to—noise ratio of 5 dB, there was no loss of intelligibility associated with the enhancement technique.

from that reported by others in two major ways: we subtract a factor (a) times the noise first, spectrum, where a is a number greater than unity and varies from frame to frame. Second,we prevent the spectral components of the processed signal from going below a certain lower bound which we call the sceotral floor. We express the spectral floor as a fraction of the original noise power spectrum Pn(w).

,

2. BASIC METHOD
The basic principle of spectral noise subtraction appears in the literature in various implementations [1_1]. Basically, most methods of speech enhancementhave in common the assumption that the power spectrum of a signal corrupted by uncorrelated noise is equal to the sum of the The signal spectrum and the noise spectrum. preceding statement is true only in the statistical sense. a However, taking this assumption as reasonable approximation for short—term (25 as) spectra, its application leads to a simple noise subtraction method. Initially, themethod we implemented consisted in computing the power spectrum of each windowed segment of speech and subtracting from it an estimate of the noise power The estimate of the noise is formed spectrum. The original phase of during periods of "silence". the OFT of the input signal is retained for the enhancement algorithm resynthesis. Thus, consists of a straightforward implementation of the followingrelationship:

1. INTRODUCTION

We report on our work to enhance the quality of speech degraded by additive white noise. Our goal is to improve the listenability of the speech signal by decreasing the background noise, without The affecting the intelligibility of the speech. noise is at such levels that the speech is essentially unintelligible out of context. We use the average segmentalsignal—to—noise ratio (SNR) to measure the noise level of the noise—corrupted speech signal. We found that sentences with a SNR in the range —5 to +5 dB have an intelligibility score in the range 20 to 80%. There is strong a correlation between the intelligibility of sentence and the SNR, but intelligibility also on context, and on the depends on the speaker, phonetic content. After an initialinvestigation of several methods of speech enhancement, we concluded that the method of spectral noise subtraction is more effective than others. In this paper we discuss our implementation of that method, which differs

let D(w) = P5(w)—P0(w)

P(w)

0,

D(w)>O ID(v), otherwise

if

(1)

where P(w) is the modified signal spectrum, P5(w) is the spectrum of the input noise—corrupted speech,...