Determination of Formant Features in Czech and Slovak for GMM Emotional Speech Classifier

Abstract

The paper is aimed at determination of formant features (FF) which describe vocal tract characteristics. It comprises analysis of the first three formant positions together with their bandwidths and the formant tilts. Subsequently, the statistical evaluation and comparison of the FF was performed. This experiment was realized with the speech material in the form of sentences of male and female speakers expressing four emotional states (joy, sadness, anger, and a neutral state) in Czech and Slovak languages. The statistical distribution of the analyzed formant frequencies and formant tilts shows good differentiation between neutral and emotional styles for both voices. Contrary to it, the values of the formant 3-dB bandwidths have no correlation with the type of the speaking style or the type of the voice. These spectral parameters together with the values of the other speech characteristics were used in the feature vector for Gaussian mixture models (GMM) emotional speech style classifier that is currently developed. The overall mean classification error rate achieves about 18 %, and the best obtained error rate is 5 % for the sadness style of the female voice. These values are acceptable in this first stage of development of the GMM classifier that should be used for evaluation of the synthetic speech quality after applied voice conversion and emotional speech style transformation.