Adding controlled amount of noise to improve recognition of compressed and spectrally distorted speech

Abstract

This paper deals with the recognition of speech whose spectrum is notably distorted by lossy compression (namely MP3) or by some implementations of 'speech enhancement' techniques. We show that these non-linear treatments can introduce gaps in spectrum that significantly change the distribution of MFCCs and degrade performance of ASR. We propose a method that measures the level of spectrum distortion and use it for adding a controlled amount of noise to the signal. It effectively masks the gaps and helps namely in situations where the source and parameters of the distortion are not known and hence we cannot use a properly matched acoustic model. In spite of its simplicity, the method can improve significantly speech recognition of highly compressed or spectrally distorted signals. We demonstrate it in several large experiments conducted on publicly available speech databases, in two languages and for two types of spectral distortion.

Common sounds in bedrooms (CSIBE) corpora for sound event recognition of domestic robots

“…The first MFCC coefficient was dropped, the remaining were added to the feature vector. There is no clear consensus in the literature about the ideal MFCC count, some earlier studies employed 13 MFCC components (Beltrán et al, 2015;Chu et al, 2009;Terence et al, 2013), but other works included 15 (Phan et al, 2015), 16 (Mesaros et al, 2010), 20 (Ruiz-Martinez et al, 2013), 26 (Salamon et al, 2014) and 40 (Nouza et al, 2013).…”

The optimization of PLP feature extraction for LVCSR recognition of MP3 data

“…The authors of [4], [5] theoretically studied the effect of MP3 distortion on the extracted features. The authors argued that the main problem of using the compressed data comes from "spectral holes" caused by the MP3 masking, which significantly alters the 2 nd and 3 rd derivatives of standard spectral-based features.…”

Advanced acoustic modelling techniques in MP3 speech recognition

“…The authors consistently reported a significant drop in accuracy for bit-rates lower than 24 kbps, i.e., [3][4][5][6][7]. Several solutions have been proposed to improve the recognition for lower bit-rates, starting with limiting the training signal bandwidth, using perceptual linear prediction (PLP) features or adding a controlled amount of noise.…”

“…If the standard procedure to avoid the Inf. values in logarithmic spectra is to add small amounts of uniformly distributed noise, then the addition of relatively strong noise has been shown to improve the recognition of spectrally distorted speech [7]. This technique is referenced as additional dithering later in the text.…”

“…In [7], the authors studied the effect of spectral valleys on speech features and concluded that the main portion of degradation could be attributed to the spectral valleys. The valleys act as a step energy change between neighboring frames, which increases the values of and 2 parameters and randomly displaces their position in the feature space.…”