Auditory Physiology Based Emotion Recognition System

This research proposes an emotion recognition system that lies on mimicking the human auditory perception model based on two-stage auditory module, including the early stage, and the cortical stage. It stimulates the output auditory spectrogram of the early stage by MFCCs feature extraction, employs Gabor filter banks as spectral-temporal analysis in the cortical stage, and embeds CNN structure in the final stage to refine and fusion pattern of features. The experimental results show that the proposed method outperforms the existing methods in speech emotion recognition.