Computational intelligence technique to predict occlusion therapy using pattern visual evoked potential P VEP

Abstract

In this thesis, we present at end-to end brain-computer interface system based on pattern visual evoked potential (P-VEP). The system uses P100 signal from the brain, a positive - evoked potential caused by flickering of checkerboard pattern. The application to analysis the role of P100 we have chosen occlusion therapy treatment which has been given for squint eye patient. P100 component can also be used to analysis the effectiveness of occlusion therapy. We have designed a visual stimulus that can be personalized to user performance. We have developed and implemented P-VEP signal processing, learning and classification algorithms. Our classifier is based on Linear Discriminate Analysis (LDA), in which we have explored choices of channel optimization using heuristic genetic algorithm and improvements. In order to predict the channel value, it has been cross validates using regression and standard regression. Data has been collected based on offline and online for decision - making. We have proposed modifications in the stimulus and decision-making procedure using Multilayer Preceptorn (MLP) to predict the vision improvement and to increase the online efficiency. We have evaluated the classification accuracy, by analyzing two different algorithms K-means and Support Vector Machine (SVM). And find that SVM can be considered as better classifier which can be used along with MLP We have evaluated the effectiveness of the system on 7 healthy subjects on visualizing and observed that the system achieves higher average speed than the system reported in the literature for a given classification accuracy. newline