Asim M. Mubeen
Candidate Ph.D. (Computational Biomedical Physics)
Cyberphysics Lab.
Department of Physics
University at Albany
Albany NY-12222
Education
- Doctor of Philosophy (Computational Biomedical Physics), Expected Sept 2013, Department of Physics, University at Albany, Albany NY, USA
- Master of Science (Physics), 2007, Department of Physics, University at Albany, Albany NY, USA
Interests
My interests includes Brain-Computer interface, Brain-Robot Interface, Source separation and localization, Bio-signal processing, Optimization theory, Computational Physics and Bayesian inference.
Research
Recently I am working on non-invasive Electroencephalography (EEG) signal processing for Brain-Computer interface and Brain-Robot Interface in collaboration with Dr. Dennis J. McFarland? at the Wadsworth Center, New York State Health Department. I am working to develop new techniques to estimate the single-trial EEG components that comprise the P300 response, which is used in a wide-variety of (BCI) paradigms. This project involves extending a specialized source separation technique called differentially Variable Component Analysis (dVCA) which performs a Bayesian estimation of the P300 response by maximizing the posterior probability with respect to its single-trial parameters such as latency, amplitude and waveshape. The dVCA algorithm provides an estimate the P300 componentry which is then used in a Bayesian evidence-based filter that can identify the P300 in the single-trial. This is accomplished by computing a probability that the P300 is present at any given point in time. My preliminary work has shown that evidence-based filter is better in accuracy than the other commonly-used filtering techniques such as the match filter (Woody filter). Such a methodology promises to increase the channel capacity of the P300-based BCI paradigms, which will increase the speed and accuracy of existing BCI systems.
I am also working to develop practical brain-machine interfaces. Non-invasive BCI methods relying on scalp electroencephalography (EEG) have achieved two and three dimensions of movement control, which is most often implemented via cursor control on a computer screen. I am working to extend these cursor-based BCI interfaces to enable subjects to directly control a robotic avatar. By using sensors on the robot I am able both to estimate the accuracy and precision with which a subject can reach various targets and to evaluate the utility of a variety of control algorithms—intelligent or otherwise. I am planning to utilize tactile sensors on the robots themselves to provide direct tactile feedback to the subject, when possible, which I expect will strengthen the interface in measurable ways.
I am also developing a source separation technique, the source separation technique can also will be very useful in localization of epileptic foci.
Invited Talks and Presentations
- “Differentially Variable Component Analysis (dVCA): Separating P300 components” Spring 2010, Department of Physics, University at Albany, Albany NY
- “Brain Robotics Interface” April 2010, New Trends in Informatics Research 2010 College of Computing and Informatics, University at Albany, Albany NY
- “Enhancing Motion Control in Brain Machine Interfaces” June 2010, Beyond Brain Machine Interface Workshop, Long Beach CA
- “Online Detection of Brain Evoked Responses” October 2010, Department of Physics, University at Albany, Albany NY
- “Detection of P300 in Single-Trial EEG using Differentially Variable Component Analysis (dVCA)” October 2010, Biomedical Engineering Society Annual Meeting 2010, Austin TX.
- “Bayesian Evidence-Based Filter: Online Detection of Evoked Brain Responses” October 2010, 5th Annual Machine Learning Symposium, New York NY.
Goals
My goals are:
- To understand the information processing in biological systems
- To apply machine learning, optimization and computational techniques for designing automated and intelligent systems
- Improve machine learning techniques by incorporating the important information available.