OFFLINE LINEAR DISCRIMINANT ANALYSIS CLASSFICATION OF TWO CLASS EEG SIGNALS

Abstract

This paper investigates the use of LDA algorithm In the EEG classification. EEG feature extraction is Implemented to reduce the dimensionality of data. The Sliding Window Technique (SWT) is used to calculate the mean within each window samples. Then, classification is done based on hyperplane technique. The LDA algorithm is described in details with all the implementation Issues. The LDA regularization is also discussed and its effects on the classification accuracy is given. In addition, both window size and channel selection effect on the accuracy is Illustrated. Results show that a window size of 150 samples, channel 3 and regularization parameter of 0.9 gives an accuracy of 0.9.