Speech Emotion Recognition using auditory features

摘要

Instinctively, speech emotion recognition (SER) has as objective to recognize emotion, based on speech signal. In this paper, we propose and evaluate a SER system based on auditory based features in order to take advantage of the characteristic to put in place a viable method. To this end, we used Gammatone Frequency Cepstral Coefficients and GammChirp Frequency Cepstral Coefficients as feature methods. They were applied to the IEMOCAP data base using Hidden Markov Models as classifier. Additionally, we tried to test the robustness of our approach by testing it in a car noisy environment. We obtained significant performance gains with the new features in both clean and noisy environment.