Relevance Vector Machine Based EEG Emotion Recognition

Personal emotions accompany us in our daily life, affecting our learning and work, therefore it is necessary to obtain better understanding of human behavior through emotional assessment. This paper proposes a method for recognizing emotions electroencephalography(EEG) based on relevance vector machine(RVM). Emotional states of two types as positive and negative were selected from a standard database of DEAP, with relevance vector machine and support vector machine(SVM) to apply classification and comparison. Experimental results show that RVM classification accuracy was 93.33% and the test run time was 0.0156s; while SVM classification accuracy was 78.67% and the test run time was 0.0211s. Compared with SVM, RVM's time complexity and test error rate are lower, and its classification performance is better.