A spatial filtering approach to environmental emotion perception based on electroencephalography.

Studies have demonstrated that visual built environments can affect the emotions of individuals, which can be recorded and investigated using electroencephalography (EEG). To study emotional intensity in adolescents exposed to different visual built environments, we proposed an EEG-based spatial filtering method using Independent Component Analysis (ICA). Specifically, to identify effective video stimuli to induce emotions, we first developed a stimulus selection strategy using the normalized valence/arousal space model. Subsequently, we designed an optimum ICA-based spatial filter by analyzing independent component-to-electrode mapping patterns in different emotional states. Based on this, EEG signals with five emotional intensities in terms of arousal and valence dimensions were linearly projected by the designed filter to extract feature parameters. Finally, we used the Support Vector Model as the classifier to recognize emotions. In the laboratory environment, the average recognition accuracy ratios for the valence and arousal dimensions were 73.35% and 68.54% (within-participant test) and 66.98% and 62.62% (between-participant test), respectively, for the 10 participants. The experimental results validated the feasibility of the proposed ICA-based spatial filtering algorithm for emotional intensity recognition.