Abdurrahman Özbeyaz's Home Page
Master Thesis (Abstract)

Least Square Support Vector Machine (LS-SVM) classification method, used to classify biopotentials signals, is a statistical learning technique. LS-SVM uses some parameters which effects classification. Selection of these parameters is important for classification.

In this study, EMG and EEG signals classified with LS-SVM and used parameters in LS-SVM selected with Particle Swarm Optimization (PSO) method.In the proposed method, EMG and EEG signals were separated as training data and testing data. EEG signals include two different group data and EMG signals include three different group data.

In this study firstly autoregressive (AR) model was used to acquire power spectrum of EMG and EEG signals. Then LS-SVM classification method applied and regularization parameter and kernel parameter used in classification were selected with PSO. Thus, the best classification results were explored. At the end of study EEG signals were classified as epileptic seizure or not and EMG signals were classified as myopatic seizure, noropatic seizure or not.

In this study, we want to provide an automatic system that will help specialists in the diagnosing process successfully. In the LS-SVM classification based on PSO, selection of suitable parameters which effect classification was applied successfully and the best result for classification was provided.