Abstract:
Analysis of Electroencephalograph (EEG) signal in children with dyslexia requires the identification of subtle changes or variations that differs it from the norm in a signal highly coupled with various artifacts and noises. A distinguishable and distinctive feature vector plays an important role to increase classification accuracy and also reduce the overall system complexity. This paper describes the feature extraction and computation of theta/beta band power ratio to be used as a feature vector for an improved classification between normal and dyslexic children, which includes poor and capable. EEG signals were acquired from 33 subjects consisting of evenly distributed number of 11 normal, 11 poor and 11 capable dyslexic children with theta and beta band power extracted using Daubechies wavelet transform of order 8. Electrodes were localized to the known learning pathway and alternate pathway for capable dyslexic of C3, C4, P3, P4, T7, T8, FC5 and FC6. Results through boxplot showed a higher theta/beta ratio of poor dyslexic than that of normal and capable dyslexic. Both capable and poor dyslexic was observed to have a larger variability throughout all 8 electrode locations. Normal subjects tend to have a theta/beta ratio of 1 to 2.5, while poor dyslexic children have an average range of 3 to 5. It can be concluded that subjects with a high theta/beta ratio of more than 2.5 during learning-related tasks could be related to a disorder, and within this study, points toward dyslexia.
