Optimized KNN classify rule for EEG based differentiation between capable dyslexic and normal children

Information on brain state and functionality could be obtained from Electroencephalograph (EEG) signal and is suitable to be used in analyzing brain disorders such as dyslexia. Our work here concerns on optimize setting in the classification of EEG signal for capable dyslexic and normal children using KNN (K-Nearest Neighbour) classifier. Discrete Wavelet Transform (DWT) with Daubechies of order 8 (DB8) was employed to extract the power band features, which were then normalized before classification. KNN distance function algorithm Euclidean, Correlation and Cosine were applied with the value of nearest neighbor being varied from 1 till 13. Classify rules using Random, Nearest and Consensus were explored to obtain optimum result. Results showed that the classifier accuracy is 100% for normal and capable dyslexic children using Euclidean with fc-value at 5 for Random and nearest rule.