Electroencephalographic Signal Based Clustering of Stimulated Emotion Using Duffing Oscillator

Duffing Oscillator is well known for its chaotic behaviour. This paper aims at clustering of emotions from the EEG response to external audio-visual stimulus used for excitation of a subject. The EEG signal corresponding to a specific emotive stimulus is used as an excitation input to a non-linear Duffing Oscillator dynamics, and the phase trajectory plots of the two state variables of the oscillator dynamics exhibits significant differences for various emotion-excitatory stimuli. Experimental investigations reveal that injection of Gaussian noise with a Signal-to-Noise Ratio as low as 25 dB retains the results of emotion clustering, indicating robustness of clustering. Further, with different pre-classified audio-visual stimulus responsible for excitation of a specific emotion, the phase-portraits obtained from EEG data of the subject have substantial similarity, indicating accuracy in clustering.