Classification of EEG Signal from Capable Dyslexic and Normal Children Using KNN

Identifying the relationship and differences of a child's brain signal in learning related activities, could assist in the development of a proper diagnostic assessment system for dyslexia and the information serve as the basis in objectively assessing the dyslexic children performance before and after an intervention programme. This paper describes the classification of EEG signal from capable dyslexic and normal children during writing word and non-word using K-nearest neighbors (KNN). In this work, discrete wavelet transform based feature extraction was employed to extract EEG signal features and the power was calculated from the decomposed EEG signals. The signal was normalized before being classified using various distance function and numerous k value to get the optimum output. Results obtained from this work showed that KNN with correlation and cosine distance functions and with k value of 7 and 9 was able to accurately classify EEG signals from capable dyslexic and normal children. The findings of this study demonstrate that the proposed feature extraction and classification approach produces high classification accuracy.