Improving BCI speller Accuracyusing modified SVM

Aided systems are an essential way of communication among disabled and healthy people. Brain-computer interfacing (BCI) allows users to communicate over brain signals. Locked-in syndrome is a neurological condition in which the subject is incapable to speak or move. In Locked-in syndrome, the person is unable to communicate in an ordinary mode. However, the individual is aware of the surroundings and moves his eyes. In order to allow subject to communicate in a Locked-in syndrome without any help, Brain computing is a viable option. This kind of communication used for the people affected by specific motor disabilities. Presented research work focuses on the communication method (On-screen speller) for Lock-in syndrome patients by retrieving their brain signals. One method of acquiring brain signals are using EEG devices. EEG signal measures the flow of charge. Neuron generates an electrochemical pulse during information exchange. The traditional methods are used for targeted and non-targeted EEG signals of classification. The presented work proposes an approach to exclude the non-targeted EEG signals to decrease the computational overhead and increasing the speed of the training cycle. For this purpose, the Support Vector Machine classifier is used. In the proposed approach, the double averaging system applied to data space, and Support Vector Machine to improve the performance. The first average applied to data space. In the data space column and row, signals are average as per sequence increment. The second average applied to support vector machine score. This prediction method is more robust. On the other hand, the presented research also contributes to reducing the number of electrodes that gives satisfactory results.