Emotional Classification of EEG Signal using Image Encoding and Deep Learning

For humans, emotions are important and play a significant role in human insight. Emotions are commonly identified by speech, facial expression and gesture. Recently Electroencephalograph (EEG) based emotion recognition have accumulated solid interest in the research community and it provides cheap, portable and reliable techniques for emotion recognition. In this work, classification of seed database having three emotions like positive, neutral and negative was performed, which was publicly available. This paper has two parts, time-series to image conversion of EEG signal and classification of emotion. First part, the data is transformed to an image that is used to analyse the EEG signal and in second part, the transformed image passes through deep learning to understand the emotions encountered during the EEG signal generation. Experimental results indicate that the scalogram of image encoding provides the best classification accuracy of 98%, compared to spectrogram and Hilbert Huang Transform (HHT) 78% and 75% classification accuracy respectively.