Dyslexia is a well-known reading disorder that involves difficulty in fluent reading, decoding and processing of words despite adequate intelligence. It is common that the reading speed of dyslexic patients is lower than their normal counterparts, because of slow letter and word processing. Eye movements in dyslexic patients are significantly different from that of normal individuals, in terms of the presence of frequent fixations and stares in the former. This work proposes a Human Computer Interactive system to assist individuals having low reading speed to increase their reading speed by the analysis of eye movements. Eye movement data for different reading speeds is recorded using a laboratory developed Electrooculogram acquisition system. From the data, Adaptive Autoregressive (AAR) parameters, Band Power Estimates and Wavelet Coefficients are extracted as signal features. Reading speeds are classified using different pattern classifiers from which an average accuracy of 94.67% over all classes and participants is obtained using Radial Basis Function (RBF) Support Vector Machine (SVM) Tree classifier and AAR Parameters as features. Friedman test is done to select the best classifier. The trained classifier is used to recognize the reading speeds of a set of new normal individuals. If the reading speeds are less than a preset threshold, that individual is trained repeatedly for 10 days for improvement. An improvement of reading speed is observed by the decrease in the misclassification rate from 45.1% to 9.92% in 10 days for the fastest speed (1 sentence/2 s) over all the subjects. This work is carried out on healthy individuals. However, the results reveal that the proposed system may also be used for training and assisting children with dyslexia or other similar reading disabilities children.