Seizure detection using wavelet transform and a new statistical feature

In this paper, we suggest a new set of statistic feature for the electroencephalogram (EEG) signals classification. We use two methods of seizure detection for evaluate new of statistic feature. Initially, features are extracted from EEG signals by using discrete wavelet transform. Next, a set of statistical features are extracted from each frequency sub-band to represent the distribution of wavelet coefficients. We suggest three new statistical features, fourth moment divided by second moment, difference between maximum and minimum, and zero-crossing of the wavelet coefficients. We demonstrate proposed features are very efficient for EEG classification and cause to improve correct classification rate (CCR). So, we use a linear discriminant analysis (LDA) and multilayer perceptron (MLP) for features selection. Next, the resultant data are applied to the classifiers. Two classifiers are employed: K-nearest neighbors (K-NN) and Bayesian. The data are classified into three categories: healthy volunteers, epilepsy patients during seizure-free interval and epilepsy patients during seizure. The experimental results indicate that performance of our method in EEG classification signals outperforms previously presented methods