Speech and Vision Lab

The objective of this paper is to find the fundamental difference between breathy and modal voices based on differences in speech production as reflected in the signal. We propose signal processing methods for analyzing the phonation in breathy voice. These methods include technique of zero-frequency filtering, loudness measurement, computation of periodic to aperiodic energy ratio and extraction of formants and their amplitudes using group-delay based technique. Parameters derived using these methods capture the excitation source characteristics which play a prominent role in deciding the voice quality. Classification of vowels into breathy or modal voice is achieved with an accuracy of 93.93% using the loudness measure