A real-time wearable emotion detection headband based on EEG measurement

A real-time emotion detection system based on electroencephalogram (EEG) measurement has been realised by means of an emotion detection headband coupled with printed signal acquisition electrodes and open source signal processing software (OpenViBE). Positive and negative emotions are the states classified and the Theta, Alpha, Beta and Gamma frequency bands are selected for the signal processing. It is found that, by using a combination of Power Spectral Density (PSD), Signal Power (SP) and Common Spatial Pattern (CSP) as the features, the highest subject-dependent accuracy (86.83%) and independent accuracy (64.73%) is achieved, when using Linear Discrimination Analysis (LDA) as the classification algorithm. The standard deviation of the results is 5.03. The electrode locations were then improved for the detection of emotion, by moving them from Fl, F2, T3 and T4 to Al, F2, F7 and F8. The subject-dependent accuracy, using the improved locations, increased to 91.75% from 86.83% and 75% of participants achieved a classification accuracy higher than 90%, compared with only 16% of participants before improving the electrode arrangement. (C) 2017 Elsevier B.V. All rights reserved.