No abstract

No abstract

Electroencephalogram (EEG) is an important technique for detecting epileptic seizures. In this paper a method of classification of EEG signal into normal, interictal and ictal classes is presented. Statistical measures such as median absolute deviation (MAD), variance and entropy showing the dispersion and rhythmicity, were calculated for each frame of EEG signals. The classification was done using a linear classifier. The direct time domain approach adopted without resorting into any kind of transformations yields an accuracy of 100%.

The statistical properties of seizure EEG are found to be different from that of the normal EEG. This paper ascertains the efficacy of inter quartile range (IQR), a median based measure of statistical dispersion, as a discriminating feature that can be used for the classification of EEG signals into normal, interictal and ictal classes. IQR along with variance and entropy are calculated for each frame of EEG. To reduce the feature vector size, standard statistical features such as mean, minimum, maximum and standard deviation were evaluated and were given as input to a linear classifier. Without resorting to any kind of transformation, the proposed method reduces the computational complexity and achieves a classification accuracy of 100%. General TermsSignal processing, pattern recognition.

Electroencephalogram (EEG) is the major diagnostic tool used for analyzing the human epileptic seizure activity and there is a strong need of an efficient automatic seizure detection using it to ease the diagnosis. In this paper a method of classification of EEG signals using wavelet based features is presented. The wavelet decomposition was done up to fourth level, followed by the calculation of inter quartile range (IQR), an important statistical feature, over third and fourth level wavelet coefficients. The methodology was applied to five types of EEG signals: healthy subjects (eyes open and eyes closed), epileptic subjects during seizure free interval (interictal EEG from epileptogenic zone and opposite hemisphere of epileptogenic zone) and epileptic subjects during a seizure (ictal EEG). A linear classifier trained on these features could classify normal and ictal EEG signals with 100% sensitivity and specificity. The overall accuracy obtained for five classes was 95.6%.