EEG signal classification in non-linear framework with filtered training data

Electroencephalographic (EEG) signals are produced in brain due to firing of the neurons. Any anomaly found in the EEG indicates abnormality associated with brain functioning. The efficacy of automated analysis of EEG depends on features chosen to represent the time series, classifier used and quality of training data. In this work, we present automated analysis of EEG time series acquired from two different groups. Non-linear features have been used here to capture the characteristics of EEG in each case since it portrays the non-linear dependencies of different parameters associated with EEG. In the first case, we present the classification between alcoholics and controls. In the second case, we present classification between epileptic and controls. In the classification, we have addressed the issue of quality of training data. In the proposed scheme prior to classification, we filter the training data. This approach led to minimum 10% improvement in the classification accuracy.