Decision tree based classification of EEG signals recorded during two dimensional cursor movement imagery

Input signals of a EEG based Brain computer interface (BCI) system are naturally non-stationary, have poor signal to noise ratio, dependent on physical or mental tasks, and contaminated with various artifacts, such as electromyogram and electrooculogram. All these disadvantages motivate the researchers substantially improve the speed and accuracy of all components of the communication system between the brain and a BCI output device. Hence, it is significant to use optimal classification algorithm and low dimensional feature set to implement a fast and accurate BCI system. On the other hand, it is very important that the classifiers have the ability for discriminating signals which are recorded in different sessions to make brain computer interfaces practical in use.In this thesis, we propose fast and accurate classification methods for classifying of up/down/right/left computer cursor movement imagery EEG data. Data sets were acquired from three healthy human subjects in age group of between 24 and 29 years old and on different days in two sessions. Extracted feature vectors based on continuous wavelet transform coefficients, autoregressive parameters, skewness and average value of derivative of the EEG signals were classified by k-nearest neighbor, support vector machine and linear discriminant analysis algorithms. The proposed methods were successfully applied to our data sets and achieved 60.53%, 62.50% and 84.21% classification accuracy rate on the test data of three subjects.