CLASSIFICATION OF DIFFERENT SPEAKING GROUPS BY MEANS OF VOICE QUALITY PARAMETERS

摘要

This paper presents a new method to classify different speaking groups by using the so called voice quality parameters. By voice quality we mean the characteristics of the glottal excitation of the speech signal. We estimate five parameters describing the voice quality as spectral gradients of the vocal tract compensated speech. They correlate with the glottal speech source features like open quotient, skewness of the glottal pulse, and incompleteness of closure. These parameters are then used to classify gender, different phonation types, and different emotions with promising results.