%Aigaion2 BibTeX export from Idiap Publications
%Saturday 22 June 2024 03:01:02 PM

         author = {Gysels, E. and Mill{\'{a}}n, Jos{\'{e}} del R. and Chiappa, Silvia and Celka, P.},
       projects = {Idiap},
          month = {11},
          title = {Studying Phase Synchrony for Classification of Mental Tasks in Brain Machine Interfaces},
      booktitle = {Proceedings of the Conference of the International Society for Brain Electromagnetic Topography},
           year = {2003},
        address = {Santa Fe, USA},
       abstract = {Electroencephalogram recordings during imagination of mental tasks allow for developing a new communication device for, e.g., motor disabled people. 32-channel EEG was recorded from 5 healthy subjects while performing, after instruction and in random order, repetitive left-hand movement imagination, right-hand movement imagination and word generation. In 3 consecutive days, 5 sessions of 4 minutes were acquired, without feedback. Extracted features were based on the Phase Locking Value (PLV,',','),
 coherence, and the spectral power in the a-, b1-, b2- and 8-30Hz frequency bands. Different feature subsets were considered for classification. The classifier consisted of a combination of Support Vector Machines (SVMs,',','),
 allowing for classifying the 3 mental tasks. Results were obtained from 5-fold crossvalidation. We analyzed (offline) data from the last day of recording, when subjects had a training experience of only 40 minutes from the two previous days. Selecting for every subject the best feature subset, often a combination of PLV and power features, we obtained for two subjects correct classification rates of 67.81\% and 62.58\%, distinguishing the 3 mental tasks. For the other subjects, classification rates were 52.77\%, 44.66\%, and 41.24\% only, maybe due to not enough practice. The involvement of the motor cortex manifested itself by a good performance of the features extracted from electrodes located in the centro-parietal region. Sole use of PLV or coherence features yielded good classification, proving phase synchronization appropriate for classifying mental tasks. This shows the importance of studying functional relations between brain regions for further improvement of Brain-Machine Interfaces.},