Atributos Acústicos Baseados na Simetria Glotal e no Classificador α -GMM para Identificação de Emoções e Locutor

This paper presents a study of acoustic emotion recognition and its effects on speaker recognition systems. Mul- tistyle emotion identification experiments were performed using a feature based on the glottal pulse symmtery, which is estimated from the source excitation signal, and compared with the results achieved by the CB-TEO-Auto-Env feature. Speaker and emotion dual identification experiments were also performed through feature fusion with Mel-frequency cepstral coefficients (MFCC). The degration on the accuracy allowed the discrimination of emotions along with the α value of the α-GMM classifier.