User Manual Privacy Policy Disclaimer Contact us
  Advanced SearchBrowse





Subject independent pattern classification of overt and covert movements from fNIRS signals


Zaidi,  AD
Department Physiology of Cognitive Processes, Max Planck Institute for Biological Cybernetics, Max Planck Society;
Max Planck Institute for Biological Cybernetics, Max Planck Society;

External Ressource

(Publisher version)

Fulltext (public)
There are no public fulltexts stored in PuRe
Supplementary Material (public)
There is no public supplementary material available

Zaidi, A., Robinson, N., Rana, M., & Sitaram, R. (2013). Subject independent pattern classification of overt and covert movements from fNIRS signals. Poster presented at 43rd Annual Meeting of the Society for Neuroscience (Neuroscience 2013), San Diego, CA, USA.

Cite as: http://hdl.handle.net/21.11116/0000-0001-4E35-D
Several studies have reported on the feasibility of using Near Infra-Red Spectroscopty (NIRS) for developing brain-computer interface (BCI) devices as an alternate mode of communication and environmental control for the disabled, including its application in neurofeedback training. In the present study, we report the development of a real-time Support Vector Machine (SVM) based pattern classification and neurofeedback system using multi-channel NIRS. We used left versus right hand movement execution and movement imageries for training and testing the classifier. Subjects performed hand movements similar to clenching a ball. We conducted three experiments to test the robustness of our classifier system, training the classifier on movement imagery and testing on movement execution, or vice versa. In the first two experiments, activations in the motor cortex during motor execution and movement imagery were used to develop subject-specific models. The classifiers implemented an adaptive bias-correction algorithm. We obtained high classification accuracies establishing the robustness of the classifier. The third experiment focused on real-time binary classification of movement execution and movement imagery using a subject-independent classifier that was trained on movement execution data from four subjects. The average online classification accuracies for subject-independent classification of new, untrained subjects were approximately 63 for movement imagery and 80 for movement execution. We also performed offline analysis to identify the spatial patterns of activation and the classifier parameters. This method of classification has applications towards rehabilitation of stroke patients that have improper activation patterns in their motor cortex. Furthermore, the capability of our system to successfully classify activation patterns from both motor imagery and motor execution supports earlier reports that their underlying neural patterns have common