CSEE FRIDAY SEMINAR SERIES – NOVEMBER 20, 2009
Flarsheim Hall 557 – 2:30 – 3:30 P.M.
Subject-Independent Brain-Computer Interfaces
Presenters: Jesse Sherwood and Reza Derakhshani
Brain Computer Interface (BCI) systems, also known as thought translation devices, permit direct communications between the user and a computer device. One of the challenges in BCI research is achieving subject independence. We introduce an ensemble-based approach for subject-independent BCI (SIBCI), and present its results on a range of fusion methods at different input and output stages. Data was collected from ten (10) randomly selected, untrained subjects using a low-end 20 channel EEG system. As untrained subjects may lack the discipline to create pure motor imagery, a four-class protocol employing slight restrained limb movements (quasi movements) was utilized. An error-correcting code was devised to map the original unbalanced four-class problem to a six-metaclass balanced dataset. A range of different EEG features (modalities) were derived and ranked based on their subject-invariant Support Vector Machine (SVM) classification results. An ensemble of SVM classifiers, each operating on one of the aforementioned input modalities and combining right and left hemisphere data, were fused through majority vote and error correcting codes using 6-fold cross-validation. Our single mode classifiers provided accuracies ranging from slightly better than chance, to 56.9%. By fusing classifier votes, our multimodal ensemble was able to achieve subject independent classification accuracy of up to 70.6% on the four quasi-movement, non-ideal test dataset;
an 18% improvement over our best single-mode, non-ensemble classifier.
Reza Derakhshani, Ph.D. is an Assistant professor at UMKC’s School of Computing and Engineering, Department of Computer Science/Electrical Engineering. His research focuses on computational intelligence with applications in biometrics and biomedical signal analysis. He earned his Ph.D. and Master’s degrees in Computer Engineering and Electrical Engineering respectively from West Virginia University. He earned his bachelor's degree in Electrical Engineering from Iran University of Science and Technology. His work has been mainly funded by the National Science Foundation, and has resulted in a number of peer-reviewed publications and a U.S. patent.
Jesse Sherwood received the MS in electrical and computer engineering from University of Missouri at Columbia, and the BSEE from University of Missouri at Rolla and holds an MBA from Rockhurst University. He is presently a Graduate Research Assistant at the Computational Intelligence and Bio-Identification Technology (CIBIT) Laboratory of the Electrical and Computer Engineering Department at the University of Missouri at Kansas City, Kansas City MO, USA where he is the recipient of the 2009-2010 UMKC Chancellors Doctoral Fellowship. He served as Chief Technical Officer at Illuminet, where his work resulted in a U.S. patent. He previously worked for Sprint, Agilent Technologies, and Tekelec. Prior to that he worked at an Engineering Consultant in Broadcasting Technology. His current research interests are in the area of Biomedical Signal Processing, Feature Extraction and Classification of EEG signals. Mr. Sherwood is a member of Eta Kappa Nu, Tau Beta Pi, Upsilon Pi Epsilon and Omicron Delta Kappa. He is a registered Professional Engineer in Missouri and Kansas.
Presenters: Jesse Sherwood and Reza Derakhshani
Brain Computer Interface (BCI) systems, also known as thought translation devices, permit direct communications between the user and a computer device. One of the challenges in BCI research is achieving subject independence. We introduce an ensemble-based approach for subject-independent BCI (SIBCI), and present its results on a range of fusion methods at different input and output stages. Data was collected from ten (10) randomly selected, untrained subjects using a low-end 20 channel EEG system. As untrained subjects may lack the discipline to create pure motor imagery, a four-class protocol employing slight restrained limb movements (quasi movements) was utilized. An error-correcting code was devised to map the original unbalanced four-class problem to a six-metaclass balanced dataset. A range of different EEG features (modalities) were derived and ranked based on their subject-invariant Support Vector Machine (SVM) classification results. An ensemble of SVM classifiers, each operating on one of the aforementioned input modalities and combining right and left hemisphere data, were fused through majority vote and error correcting codes using 6-fold cross-validation. Our single mode classifiers provided accuracies ranging from slightly better than chance, to 56.9%. By fusing classifier votes, our multimodal ensemble was able to achieve subject independent classification accuracy of up to 70.6% on the four quasi-movement, non-ideal test dataset;
an 18% improvement over our best single-mode, non-ensemble classifier.
Reza Derakhshani, Ph.D. is an Assistant professor at UMKC’s School of Computing and Engineering, Department of Computer Science/Electrical Engineering. His research focuses on computational intelligence with applications in biometrics and biomedical signal analysis. He earned his Ph.D. and Master’s degrees in Computer Engineering and Electrical Engineering respectively from West Virginia University. He earned his bachelor's degree in Electrical Engineering from Iran University of Science and Technology. His work has been mainly funded by the National Science Foundation, and has resulted in a number of peer-reviewed publications and a U.S. patent.
Jesse Sherwood received the MS in electrical and computer engineering from University of Missouri at Columbia, and the BSEE from University of Missouri at Rolla and holds an MBA from Rockhurst University. He is presently a Graduate Research Assistant at the Computational Intelligence and Bio-Identification Technology (CIBIT) Laboratory of the Electrical and Computer Engineering Department at the University of Missouri at Kansas City, Kansas City MO, USA where he is the recipient of the 2009-2010 UMKC Chancellors Doctoral Fellowship. He served as Chief Technical Officer at Illuminet, where his work resulted in a U.S. patent. He previously worked for Sprint, Agilent Technologies, and Tekelec. Prior to that he worked at an Engineering Consultant in Broadcasting Technology. His current research interests are in the area of Biomedical Signal Processing, Feature Extraction and Classification of EEG signals. Mr. Sherwood is a member of Eta Kappa Nu, Tau Beta Pi, Upsilon Pi Epsilon and Omicron Delta Kappa. He is a registered Professional Engineer in Missouri and Kansas.