Emotion Classification of Infant Voice Based on Features Derived from Teager Energy Operator

To study effective speech features which can represent different emotion styles in infant voice, nonlinear features based on Teager Energy Operator are investigated. Neutral state and 4 emotional states (i.e. happiness, impatience, anger and fear) are classified from the infant voice database. MFCC extraction and HMM-based emotion classification are used as baseline system to evaluate the emotional classification performance of nonlinear features. In comparison with MFCC, relative improvements which are 2%, 2% , 2% and 10% of classification capacity are obtained when using NFD_Mel , AF_Mel, DAF_Mel and TEO_SBCC. But the performance of emotion classification decreases respectively by 14% for using AM_SBCC.