ECoG-based of finger movement classification with KNN and SVM

In this thesis, two approaches have been followed classification of finger movements have aimed using ECoG recordings. In first approach, performance in classification of clusters are determined with k-NN and SVM classifiers by using AR coefficients as attribute. According to found results, it has been seen that maximum performance is obtained. With less coefficients generally. From the perspective of the classifier performance of SVM is beter than that of k-NN. In second approaches followed for classification the values of ECoG, wavelet coefficients are used as an attribute. in the second stage only SVM classifier is used because it was found firstly that SVM performance was beter than that of k-NN. It is seen that classification performance rates are increased by determining of effective channels. It has been tried to understanding the dynamics of brain by evaluated EEG / ECoG signals obtained with different paradigms. In this study ECoG signals are used for the same aim. It is seen that these methods can be applied successfully.