Page - 156 - in Intelligent Environments 2019 - Workshop Proceedings of the 15th International Conference on Intelligent Environments

Table 1. DEAP data format Array name Array shape Array contents data 40 x 40 x 8064 video/trial x channel x data labels 40 x 4 video/trial x label (valence, arousal, dominance, liking) DEAP uses an EEG with 32 electrodes. Figure 4 shows the electrode arrangement of the EEG used in the experiment. Figure 4. Electrode position in 10-20 method Similar signals can be obtained from adjacent electrodes, but the signals acquired from the distant electrode are not similar as the nearer electrodes. This motivates us to investigate the selection of effective electrodes. 2.2.2 Division of EEG Since the EEG signals were taken at the time of watching a 60-s moving image, each data is composed of 60 seconds. Candra and colleagues examined the appropriate window size in recognition of Valence and Arousal's emotions using DEAP [16], which suggested that it would be better to divide the electroencephalogram data every 6 seconds and proceed with the analysis. In this study, based on this result, the EEG was divided every 6 seconds and analyzed. Although the original signal was sampled at 128 Hz, it was separated into a γ frequency band of 33 to 64 Hz, a β frequency band of 17 to 32 Hz and an α frequency band of 9 to 16 Hz by wavelet transform as follows. ∑ ∑ (2) 2.2.3 Calculation of energy By the processing so far, the EEG was divided every 6 seconds, and the EEG of each segment was separated into three frequency bands γ, β, α. In order to investigate how much EEGs of each separated frequency band are contained, energy of each frequency band was calculated using Equation 3. ∑ (3) Y.Tsurugasaki etal. / IdentificationofEffectiveEEGElectrodes forDepressionSensing156