Analysis of adaboost classifier from compressed EEG features for epilepsy detection

Epilepsy is one of the serious and chronic neurological disorders affecting millions of people in the world. The seizures produced due to epilepsy manifests itself in various forms of symptoms ranging from a short lapse of attention to severe muscle jerks. To diagnose whether a patient is epileptic or non epileptic and to determine the required treatment is the most important stage when dealing with the patient. In the present scenario, diagnosing such disorders is executed manually by neurologists whose availability is limited. With the advent of computer based diagnosis dependent on EEG, the diagnosis could be made quite faster. For the computer based diagnosis, various signal processing algorithms and machine learning techniques play a vital role. An immense amount of information about the brain can be understood with the help of Electroencephalography (EEG) signals. An EEG signal is non-stationary and highly time varying in nature and so it can be analyzed by non-linear techniques. In this paper, the concept of code converters is implemented and as the classification results through it are not satisfactory it is further optimized with the help of Adaboost Classifier. Results show that when Adaboost Classifier is utilized as a post classifier, an average perfect classification rate of about 94.58%, an average classification accuracy of about 97.29%, an average performance index of about 94.51 % and an average quality value of 21.82 is obtained.