9th Annual Conference of the International FES Society September 2004 – Bournemouth, UK
Communication speed enhancement for visual based Brain Computer Interfaces Sami S, Nielsen KD Centre for Sensory Motor Interaction (SMI), Aalborg University, Denmark Email:
[email protected]
Abstract This study aimed at investigating the feasibility of enhancement of the speed of current Steady State Visual Evoked Potential (SSVEP) based Brain Computer interfaces. We used infomax independent component analysis (ICA) to decompose 6-channel EEG data from trials in which the subjects gazed at an 8.8 Hz and 35Hz flashing flicker blocks simultaneously. Applying time-frequency analysis to the time courses of activity of the resulting six independent EEG components revealed that successful regulation of the measured activity was accompanied by significant changes in power and phase coherence in the occipital cortex. Results indicate that responses within the range of one second could be used to further optimize the performance of visual based braincomputer interfaces.
1
Introduction
The concept of a Brain Computer Interface (BCI) has emerged from the need for alternative communication and control options for individuals with severe motor disabilities such as amyotrophic lateral sclerosis or cerebral palsy. The potential usage of such an interface extends to rehabilitation of neurological disorders, brain-state monitoring as well as non-medical applications. From a biomedical perspective this kind of interface could increase the disabled individuals ability to interact with his environment, and may lead to an improved quality of life. In recent years several research approaches have been developed [1][2][3]. The most practical and widely applicable BCI solutions are those based on non-invasive electroencephalogram (EEG) measurements recorded from the scalp. The EEG has a advantage of being significantly cheaper and more accessible to the users than other neurophysiological techniques. It is therefore
seen that the EEG is a good basis for such communication and/or control. The approach presented here makes use of EEG responses known as steady state visual evoked potentials (SSVEP). SSVEPs are the brain’s electrical responses to visual stimuli [4]. It has been demonstrated that steady state visual evoked potentials (SSVEP) can be used in subjects with an intact visual sensory system for the purpose of communication [8]. The aim of the present study was to investigate the feasibility of early detection of these SSVEP responses by the use of an independent component analysis (ICA) “infomax” algorithm. Thereby investigating the possibility of further speed enhancement of already existing SSVEP based BCI systems [8].
2
Methods and Materials
In the main experiment eight subjects were placed 50cm in front of a Nokia 445X pro CRT computer screen. Rectangular blocks with a size of 60x70mm2 were displayed. The stimulation consisted of two flashing blocks flickering at 8.8Hz and 35Hz, which were presented simultaneously 12cm apart. The subjects were asked to look at the cross on the centre of the screen while they mentally focused on one of the two blocks as instructed. The sets of experiments were repeated 4 times with each subject and each trial lasted 1min. EEG was recorded from frontal, central and occipital locations using a NeuroscanTm system and a 1000 Hz sampling freq. Using this simple paradigm with flashing symbols on the stimulator screen we have performed experiments where the objectives were: To determine which ICA components are behaviorally relevant and should be selected for further investigation with usage of the
9th Annual Conference of the International FES Society September 2004 – Bournemouth, UK
EEGLAB software[7]. ICA has proven capable of isolating both artifactual and neurally generated EEG sources [9]. After identification of the relevant components further analysis was performed by deploying Inter-trial coherence (ITC). ITC is a frequencydomain measure of the partial or exact synchronization of activity at a particular latency and frequency to a set of experimental events to which EEG data trials are time locked.[9]
Figure 2 shows the early detection of the 35Hz response in one subject represented in both the event related power spectrum as well as the inter-trial coherence with bootstrap significance level: p