Multi-feature Fusion Speech Emotion Recognition Based on SVM

Abstract

With the rapid development of the computer industry, artificial intelligence gradually enters into people’s life, which makes people put forward higher requirements for human-computer interaction in the intelligent car cockpit. In order to better meet the actual requirements of emotion recognition in the cockpit, this paper proposed a multi-feature fusion speech emotion recognition system based on Gaussian kernel nonlinear support vector machine (SVM), it is suitable for voice human-computer interaction systems. In the proposed system, four features are extracted and fuse them based on weighted sequence feature fusion algorithm, then use SVM to classify six emotions and evaluated the model under Berlin Emotion Dataset(EmO-DB). Experimental results show that the recognition model with Mel frequency ceptrum coefficient (MFCC) as the main feature has high accuracy and stability.