Emotion Detection Through EEG Signals Using FFT and Machine Learning Techniques

Abstract

Detecting human emotions using various verbal and non-verbal communicating means such as facial, speech, textual and physiological signals are in use from past long time. However, verbal and facial methods are prone to deception. Therefore, emotion recognition through physiological signals has become an active area of research nowadays. Further, Brain–Computer Interaction (BCI) and Emotion detection through Electroencephalographic (EEG) signals are popular domains for the detection of emotional states, whereas the accurate processing of the EEG signals is still a major challenge due to noise and different bands of frequencies present in the signals. In the presented exposition, a methodology using the combination of Fourier Fast Transform (FFT), Principal Component Analysis (PCA) and k Nearest Neighbor (k-NN) has been proposed for emotion detection. To the best of our knowledge the above mentioned combination is first time introduced in the presented paper. Fourier Fast Transform (FFT) converts a signal from a time or space domain into frequency domain and thus eases the analysis of a given signal. PCA was used for feature selection and further, the selected features were categorized into given emotional states using k-NN classifier. The presented analysis carried out using FFT outperformed the previous experiments, with an accuracy of 96.22%.