Brain Computer Interface Analysis using Wavelet Transforms and Auto Regressive Coefficients

The idea of an EEG based BCI is to assist the people unable to communicate their thoughts due to neuromuscular disorders and hence affected by motor disabilities. The BCI helps them acting as an interface between the human mind and the computer. In this paper an offline analysis of the EEG data recorded from the C3 and C4 electrodes pertaining to motor activities was done. The data obtained was preprocessed with techniques like wavelet transform and linear predictive coding was applied to it to determine the auto regressive coefficients which are treated as feature vectors to train an artificial neural network for appropriate classification. The trained net was then subjected to testing of data from 140 random trials that were taken and the accuracy was determined. The efficiency of this approach was found to be 71.5%.