Performance Comparison of TEP and VEP Responses using Bispectral Estimation to Command an Intelligent Robot Chair with Communication Aid

This paper describes the classification of navigational tasks to command a navigation system incorporated with a communication device using thought and visually evoked potentials. To develop a navigation system with communication aid for the neuromuscular disorder community, simple protocol using TEP and VEP responses has been introduced in this research work. The developed protocol has seven basic tasks such as forward, left, right, yes, no, help and relax; these basic seven task are used to control the wheel chair navigation system and also perform voice communication using an oddball paradigm. The proposed system records the brain wave signals using a wireless EEG amplifier from ten subjects while the subjects were imagining and visualizing the seven different visual tasks. For each subject, the recorded brain wave signals are pre-processed to extract the six Electroencephalography rhythmic activities and segmented into frames of equal samples. Then, this study presents the higher order spectra based features to categorize the TEP and VEP tasks using bispectrum estimation algorithm. Further, statistical features such as the mean and entropy of the bispectral magnitude are extracted and formed as a feature set. To develop a customized classification system for individual responses, the extracted feature sets are classified using Multi layer neural networks and from the results it is observed that the entropy of bispectral magnitude feature using VEP based NN model has the maximum classification accuracy of 99.29% and the mean of bispectral magnitude feature using TEP based NN model has the minimum classification accuracy of 72.14%.