Emotion recognition has been an interesting new research topic due to its various possible applications. In this paper, we proposed and evaluated various feature extraction techniques based on discrete wavelet transform (DWT) and classifiers based on artificial neural networks (ANN), for an emotion recognition system based on EEG signals. In this study, a benchmark database was used, namely DEAP database. EEG signals were decomposed using a discrete wavelet transform (DWT) with mother wavelet Daubechies 4 and extracted using three kinds of feature extractions, i.e., relative wavelet energy (RWE), wavelet entropy (WE), and Higuchi Fractal Dimension (HFD). RWE feature after DWT 5 level decomposition produced the best recognition result. Since shorter signal length is desired in a real time emotion recognition, then in this paper we evaluated several signal window lengths and found that a 10 s window size with a 30 s delay was sufficient to do the recognition. As for the ANN as a classifier, it was found that the recognition rate of the Probabilistic Neural Networks (PNN), was slightly better compared with that of the Back Propagation Neural Networks (BPNN).
- Emotion recognition