Computational Intelligence Approaches to Biosignal Processing for Motor Imagery-based Non-Invasive Brain-Computer Interfaces
Date: 10/2/10
Venue: MS020
Time: 2.30 pm
Speaker: Dr. Damien Coyle
Affiliation:
Intelligent Systems Research Centre, University of Ulster
Computational Intelligence Approaches to Biosignal Processing for Motor Imagery-based Non-Invasive Brain-Computer Interfaces
By Dr. Damien Coyle
Intelligent Systems Research Centre, University of Ulster
Abstract
A brain computer interface (BCI) allows users to communicate without movement. BCIs infer user intent through direct measures of brain activity, usually via EEGs. BCIs are the only means of communication possible for some severely disabled users and are becoming increasingly useful to healthy subjects. BCI research at the Intelligent Systems Research Centre, University of Ulster is focused around developing signal processing tools for motor imagery-based continuous control strategies in synchronous and self-paced BCI applications. The main emphasis is on two-class BCI development with recent research also aimed at multiclass systems. To improve the performance of motor imagery based BCI a major focus has been on developing a multistage signal processing framework focused around the neural-time-series-prediction (NTSPP) framework which permits multiple-step-ahead prediction of the EEG time-series, where different prediction models are trained to specialize in predicting different EEG signals. Due to network specialization on specific motor imagery tasks and EEG channels, features extracted from the predicted signals are more separable and thus easier to classify. More recently, NTSPP has been merged with the popular Common Spatial Patterns (CSP) approach which constructs linear spatial filters that maximize the ratio of class-conditional variances of EEG sources. Combining NTSPP with CSP along with a further spectral filtering (NTSPP-SF-CSP) has shown significant potential for two-class and multiclass BCIs involving hand, foot and tongue motor imagery classification. A novel post-processing module which accounts for biased behavior and improves feedback stability via control signal smoothing through Savitzky-Golay filtering to improve the online performance has also been developed. In developing this BCI a major focus is on maintaining practicality and efficient training times therefore the complete system has been developed to be quickly and easily adapted to each individual using computational intelligence based self-organizing parameter optimization techniques involving fuzzy neural networks (FNNs) and particle swarm optimization (PSO). A range of offline results and online analysis of subjects using the BCI to control games illustrate the potential for this BCI.
Short Biography
Dr. Damien Coyle received a first class degree in computing and electronic engineering in 2002 and a doctorate in Intelligent Systems Engineering in 2006 from the University of Ulster where he is now a lecturer at the School of Computing and Intelligent Systems and a member of the Intelligent Systems Research Centre. His research and development interests include biosignal processing, bio-inspired cognitive, adaptive systems and brain-computer interface technology. More recently he has been investigating computational models of neural systems and neurodegeneration related to Alzheimer's disease (AD). Dr. Coyle chairs the UKRI chapter of the IEEE Computational intelligence Society.
Further details on Dr. Coyle's research can be found here
