A hierarchical approach of Speech Emotion Recognition based on entropy of enhanced wavelet coefficients

This paper presents a hierarchical Speech Emotion Recognition method, where the speaker-independent emotional features are derived from the Teager energy (TE) operated wavelet coefficients of speech signal. The detail as well as approximate Wavelet coefficients enhanced by TE operation is used to compute entropy. Entropy values of TE operated detail and approximate wavelet coefficients reduce feature dimension, forming an effective feature vector for distinguishing different emotions when fed to a Euclidean distance based classifier in a hierarchical process. Detail simulations are carried on EMO-DB German speech emotion database containing four class emotions, such as angry, happy, sad and neutral. Simulation results show that the proposed hierarchical emotion recognition method gives quite satisfactory four-class emotion recognition performance, yet demonstrates a significant increase in versatility through its propensity for speaker independence with lower computation.