Selection of Appropriate Statistical Features of EEG Signals for Detection of Parkinson’s Disease

Analysis of signal transmission activities of human brain can give fruitful information about its functions. These information are of very importance in detection and diagnosis of different types of neurological disorders. Besides low spatial sensitivity, Electroencephalogram(EEG) signals are used for functional analysis of activities of brain because of the large temporal resolution of it. Identification of an appropriate feature of the EEG used to have a key role for its analysis. This work specifically describes feature extraction of EEG signals of persons with Parkinson's Disease(PD) by using statistical methods. Mean, standard deviation, energy, kurtosis and skewness are the five statistical features selected for this work. In addition to the extraction of features, comparative analysis of these features are also provided considering the EEGs of both normal (Non PD) and the persons with PD symptoms by using T-test. With the use of T-test, without the application of any classification techniques, the features of any two classes can be well differentiated. In the proposed approach, the results show that the p-assessment of the T-experiment is less than 0.05 and hence it can be considered that the features of the two classes are having less than 5% similarity. This fulfils the objective of detecting PD most efficiently. Out of the five features considered, Mean and Energy are the features, which are capable of differentiating the two categories of the subjects most significantly.