Objective and Subjective Assessment of the Quality and Intelligibility of Noised Speech

Abstract

Comparison of the behavior of subjective assessments of quality and intelligibility of speech distorted by white, pink and brown noise is performed. The results obtained indicate the similarity of this behavior in the range of signal-to-noise (SNR) values 0 <; SNR <; 5<;snr<;5 db,="" with="" the="" best="" quality="" and="" intelligibility="" for="" brown="" noise,="" worst="" pink="" one.="" study="" of="" behavior="" assessments="" fi="" ve="" objective="" speech="" measures="" allowed="" identifying="" features="" that="" impose="" certain="" restrictions="" on="" application="" these="" measures.="" it="" is="" shown="" segmental="" signal-to-noise="" ratio="" (ssnr),="" unlike="" other="" measures,="" insensitive="" to="" noise="" spectrum="" shape.="" logarithmic="" spectral="" distortion="" (lsd)="" bark="" -spectral="" (bsd)="" only="" partially="" consistent="" results="" subjective="" assessment.="" perceptual="" evaluation="" (pesq)="" measure="" in="" good="" agreement="" estimates="" at="" snr=""> dB, <;/snr<;5>with the best quality and intelligibility for brown noise, and the worst for pink one. The study of the behavior of the assessments of the five objective speech quality measures allowed identifying features that impose certain restrictions on the application of these measures. It is shown that segmental signal-to-noise ratio (SSNR), unlike other measures, is insensitive to the noise spectrum shape. The behavior of logarithmic spectral distortion (LSD) and bark-spectral distortion (BSD) measures is only partially consistent with the results of the speech quality and intelligibility subjective assessment. The behavior of the perceptual evaluation of speech quality (PESQ) measure is in good agreement with subjective estimates at SNR > 0 dB, whereas the tendency for such agreement is observed only for SNR >10 dB for frequency weighted segmental SNR (fwSNR) measure.