Towards enhanced information transfer rate: a comparative study based on classification techniques

Abstract

ABSTRACT One of the most important performance parameters for Assistive Devices (AD’s) based on Brain Computer Interfaces (BCIs) is the Information Transfer Rate (ITR). This study compares a hybrid BCI to a Steady State Visually Evoked Potential (SSVEP) based BCI, along with a comparison of classification techniques used for translation of user intentions. The hybrid BCI paradigm combined SSVEP & P300, where SSVEP decodes user intentions and P300 is used for Time Division Multiplexing (TDM). The classification protocols were categorised as single-step supervised classifiers and two-step unsupervised classifier. It was observed that the classification accuracy for translation of human intentions for the traditional SSVEP paradigm (93.78%) was higher than the hybrid BCI (90.76%) proposed, but still the hybrid BCI is paradigm option for development of ADs (high ITR of 81.10 bits/minute). The study compared the two classification protocols using the statistical t-value test, which concluded that (99.9% confidence level) the mean classification accuracy and mean ITR were greater for the single-step supervised classification and also that the mean FAR was lower for the single-step supervised classification. The proposed hybrid BCI with single-step supervised learning classification protocol emerged as best BCI option for the development of AD’s.