BCI oriented EEG analysis using log energy entropy of wavelet packets. (July 2018)

• Record Type: Journal Article
• Authors: Göksu, Hüseyin
• Is Part Of: Biomedical signal processing and control. Volume 44(2018)
• Page Start: 101
• Page End: 109
• Publication Date: 2018-07
• Subjects: BCI -- EEG -- Classification -- Wavelet packet analysis -- Log energy entropy -- Multilayer perceptron; Signal processing -- Periodicals; Biomedical engineering -- Periodicals; Signal Processing, Computer-Assisted -- Periodicals; Image Processing, Computer-Assisted -- Periodicals; Biomedical Engineering -- Periodicals; 610.28
• DOI: 10.1016/j.bspc.2018.04.002
• Languages: English
• ISSNs: 1746-8094

Abstract

Highlights: A novel BCI related EEG analysis method is proposed for self regulation of SCPs. Wavelet packet analysis along with log energy entropy is used. Classification is performed with a multilayer perceptron. Method is tested on BCI Competition 2003 datasets Ia and Ib. The proposed method outperforms the other published results on the same datasets. Abstract: Brain computer interfacing (BCI) is the practice of establishing a direct communication between a human and a computer by self-regulation of some electrical activity in the brain. Electroencephalography (EEG) based BCI is one of the most promising form of this application where a subject learns to control brain waves of type μ-rhythm (8–14 Hz) or β-rhythm (14–30 Hz) or P300 event related potential or slow cortical potentials (SCPs). Intents of the person is extracted using signal processing techniques. In this study self-regulation of SCPs is studied using wavelet packet analysis (WPA), rather than traditional time or frequency domain methods. WPA, by providing arbitrary time-frequency resolution, enables analyzing signals of stationary and non-stationary nature. It has better time representation than Fourier analysis and better high frequency resolution than Wavelet analysis. WPA subimages are analyzed further by log energy entropy. These feature vectors are fed into a multilayer perceptron (MLP) for classification. We use the BCI Competition 2003 datasets Ia and Ib to test our approach. Performance of the MLP is compared with that of k-NN and SVM. Our method achieves accuracies of 92.8% and 63.3% on datasets Ia and Ib respectively. This outperforms the other reported results on the same datasets including the winners of the competition. Overall, WPA followed by Log Energy Entropy and MLP provides an efficient tool for the aforementioned kind of BCI oriented EEG analysis.