HNR extraction in voiced speech, oriented towards voice quality analysis

This study tests three methods (algorithms of G. de Krom, C. d'Alessandro et al. and P. Boersma) to estimate the Harmonics-to-Noise Ratio (HNR) in speech. Tests are made on two databases of naturally connected speech designed for voice quality analysis. First, results of the three methods are compared, then the relevance of each method is analysed separately. The conclusion is that they are all good indicators of the amount of noise in speech, and though their accuracy is limited, they are efficient for voice quality analysis.