EEG-Based Emotion Recognition Using Statistical Measures and Auto-Regressive Modeling

In this paper, a novel approach towards classification of various human emotions based on statistically weighed autoregressive modeling of Electroencephalogram (EEG) signals is discussed. The proposed algorithm was proven to be superior to many related works, in distinguishing different emotions such as happiness, fear, sadness etc. The findings discussed are based on the results obtained using benchmark emotion based EEG database called DEAP. In this work, epochs were extracted from data using statistical measures such as Shannon Entropy and higher order auto-regressive model was fit to the extracted features. The model was used for classifying human emotions by feeding it into a multi-class Support Vector Machine (MCSVM). The proposed algorithm is proven to be more efficient than existing algorithms as a classification accuracy of 94.097% was obtained.