EGG Signal Multi-scale Product for Characterizing Voice Quality

This work deals with the measures of the open quotient and the fundamental frequency from the electroglottographic (EGG) signal for voice quality characterization. The parameter measurement is based on the glottal closure and opening instant detection using the EGG signal multi-***scale product. The EGG multi-***scale product consists of multiplying the wavelet transform coefficients at three successive dyadic scales. For this aim, we have used the Voqual ‘03 database constituted by audio sound files with their corresponding EGG signals pronounced by one male speaker with various voice qualities. The analysis of the fundamental frequency and its variation versus the vocal quality reveals 3 classes: the low fundamental frequency class, the middle fundamental frequency class and finally the high fundamental frequency class. The second voice source parameter which is the open quotient allows us to separate the first class into two groups.