A Hybrid ICA-Wavelet Transform for Automated Artefact Removal in EEG-based Emotion Recognition

Removing artefacts from electroencephalographic (EEG) recordings normally increases their low signal-to-noise ratio and enables more reliable interpretation of brain activity. In this paper we present an evaluation of an automatic independent component analysis (ICA) procedure, a hybrid ICA - wavelet transform technique (ICA-W), for artefact removal from EEG correlated to emotional-state. Spectral and statistical features were classified with support vector machines (SVM) to assess the performance of ICA-W against the regular ICA, in terms of the accuracy of classifying emotional states from EEG. Accuracies on data from 14 subjects are reported and the results indicate that ICA-W performs better than traditional ICA in statistical and wavelet based features whilst the best overall performance is achieved when combining ICA-W with statistical features with an average accuracy across subjects of 74.11% for classifying four categories of emotion. ICA-W is therefore demonstrated to enhance EEG-based emotion recognition applications in terms of performance and ease of application.