EEG based automatic emotion recognition using EMD and Random forest classifier

Abstract

Electroencephalogram (EEG) signal based automatic emotion recognition is an inter-disciplinary field which has been studied for a long time. In this study, empirical mode decomposition (EMD) is used for EEG-based automatic emotion classification. The EMD is used for decomposition of non-stationary and nonlinear EEG signals into different modes called intrinsic mode functions (IMFs). From the resulting IMFs different nonlinear features namely Hjorth parameters, Shannon entropy, collision entropy, differential entropy, and Higuchis fractal dimension are extracted in order to determine the characteristics associated with emotional states. The obtained features are smoothed using moving average filter with a window size of 5 samples for removing unwanted rapid fluctuations in the extracted features. Finally, the processed features are fed to random forest (RF) classifier and the accuracies are noted for different profiles of 4, 6 and 12 channels. The proposed method is applied to SJTU emotion EEG dataset (SEED). We have obtained highest classification accuracies of 89.59%, 91.45%, 93.87% using activity feature for the above 3 channel profiles in classifying positive, neutral and negative emotions respectively. Finally a comparative analysis is performed in terms of computed classification accuracies with other above mentioned features. As compared to existing methods, the proposed method has shown better performance and can be considered for recognition of human emotional states.

Cite this work

No public copy is stored for this record, so only the metadata is cited.

Stored only when both papers are in the corpus — 45 citations exist upstream.

Papers that cite the same references — the relational fallback while embeddings are still backfilling.