All you have to do filter the signal using a 1-D gabor filter or use 'dwt' for wavelet transform function in MALTAB. You can extract features from these processed signals (in case of Gabor filter) or extract features from the detail and approximation coefficients (in case of wavelet decomposition). These features can be given to SVM classifiy function after appropriate segregation of test and training data. You can refer to my thesis work : MATLAB analysis of EEG signals for diagnosis of epileptic seizures