Web Browsing using brain signals

Abstract

Various neurological diseases severely limit affected patients with the inability to express themselves through verbal and nonverbal communication. In attempts to provide “lockedin” patients the ability to express themselves, numerous strategies have been developed to use electroencephalographic (EEG) activity or other electrophysiological signals as a basis for realising a non-muscular communication channel, formally termed as a brain-computer interface (BCI) or a human machine interface (HMI). This study concerned the use of steady state visual evoked potentials (SSVEPs), which are EEG signals observed in the occipital area of the brain while a subject focuses on a flashing stimulus flickering with a particular frequency. A functional SSVEP-based brain-controlled music player was developed by R.Zerafa [1]. Data collected from training sessions of several subjects using the music player was used in this dissertation to determine whether appropriate signal processing modalities can be applied to operate a BCI without any prior training. After a thorough review of popular signal processing techniques used in research towards SSVEP-based BCIs, a multivariate statistical technique called canonical correlation analysis (CCA), which is renowned for realising high performance, calibration-less BCIs, was used for signal pre-processing and feature extraction on the provided data. Additionally, analysis included the use of a feature extraction technique adopted by R.Zerafa for the music player, called power spectral density analysis (PSDA). Two different classifiers were tested on features extracted by the said techniques, and results have shown that with no prior training, CCA achieves comparable results to those obtained by PSDA. Finally, a communication link was created between the signal processing block and a web browser. A web page was created, displaying several links to other web pages, thereby allowing the classified commands to be realised in four simple commands of “scroll-up”, “scroll-down”, “go-to” and “go back”. Due to a limited amount of time, CCA has not been thoroughly tested in real time with this application. However, successful implementation of CCA with the web-browser application would be a one-step forward towards a “plug and play” BCI.