Abstract
The aim of the study was to select the best electroencephalogram features and channel locations for defection of wrist movement intentions. The detected intentions can be used in brain-computer interfaces (BCIs) either for direct control of an artificial or virtual hand, or they can be used as an underlying binary code for execution of other tasks. 28 channel EEG was recorded while a subject performed wrist movements in four directions. Four basic feature types were extracted in the time and frequency domains for each channel following optimized filtering of the signals. The signals were split into planning and execution segments, respectively. Various delays and anticipation lengths were taken into account for each of the features, thus totaling 93 different features. The potential performance of each feature and channel for use in the classification of the EEG signals was analyzed by estimating the relative class overlap using the Davies-Bouldin Index (DBI), a widely used measure for estimating cluster separation. The best feature/channel configurations contained both channels that were close and channels that were far from motor areas. A statistical test using the channel/feature configurations that yielded the lowest 5% DBI values for motor and for non-motor channels yielded no significant difference (α= 0.05) between these two channel populations. The scope and depth of the study was limited. Plus, important parts of the signal had to be discarded to rule out interference stemming from saccadic eye movement However, our results do suggest more attention should be paid to non-motor areas in earlinked EEG data even when investigating movement related BCIs.
Language | English |
---|---|
Title of host publication | 2004 IEEE Conference on Cybernetics and Intelligent Systems |
Publisher | IEEE |
Pages | 942-946 |
Number of pages | 5 |
ISBN (Print) | 0780386442, 9780780386440 |
Publication status | Published - 1 Dec 2004 |
Event | 2004 IEEE Conference on Cybernetics and Intelligent Systems - , Singapore Duration: 1 Dec 2004 → 3 Dec 2004 |
Conference
Conference | 2004 IEEE Conference on Cybernetics and Intelligent Systems |
---|---|
Country | Singapore |
Period | 1/12/04 → 3/12/04 |
Fingerprint
Keywords
- brain-computer interface
- cluster separation
- EEG
- movement related potentials
- electroencephalography
- steady-state visual
- demodulation
- frequency domain analysis
- signal processing
- neural nets
- pattern recognition
Cite this
}
Cluster separation index suggests usefulness of non-motor EEC channels in detecting wRist movement direction intention. / Sepulveda, F.; Meckes, M.; Conway, B. A.
2004 IEEE Conference on Cybernetics and Intelligent Systems. IEEE, 2004. p. 942-946.Research output: Chapter in Book/Report/Conference proceeding › Conference contribution book
TY - GEN
T1 - Cluster separation index suggests usefulness of non-motor EEC channels in detecting wRist movement direction intention
AU - Sepulveda, F.
AU - Meckes, M.
AU - Conway, B. A.
PY - 2004/12/1
Y1 - 2004/12/1
N2 - The aim of the study was to select the best electroencephalogram features and channel locations for defection of wrist movement intentions. The detected intentions can be used in brain-computer interfaces (BCIs) either for direct control of an artificial or virtual hand, or they can be used as an underlying binary code for execution of other tasks. 28 channel EEG was recorded while a subject performed wrist movements in four directions. Four basic feature types were extracted in the time and frequency domains for each channel following optimized filtering of the signals. The signals were split into planning and execution segments, respectively. Various delays and anticipation lengths were taken into account for each of the features, thus totaling 93 different features. The potential performance of each feature and channel for use in the classification of the EEG signals was analyzed by estimating the relative class overlap using the Davies-Bouldin Index (DBI), a widely used measure for estimating cluster separation. The best feature/channel configurations contained both channels that were close and channels that were far from motor areas. A statistical test using the channel/feature configurations that yielded the lowest 5% DBI values for motor and for non-motor channels yielded no significant difference (α= 0.05) between these two channel populations. The scope and depth of the study was limited. Plus, important parts of the signal had to be discarded to rule out interference stemming from saccadic eye movement However, our results do suggest more attention should be paid to non-motor areas in earlinked EEG data even when investigating movement related BCIs.
AB - The aim of the study was to select the best electroencephalogram features and channel locations for defection of wrist movement intentions. The detected intentions can be used in brain-computer interfaces (BCIs) either for direct control of an artificial or virtual hand, or they can be used as an underlying binary code for execution of other tasks. 28 channel EEG was recorded while a subject performed wrist movements in four directions. Four basic feature types were extracted in the time and frequency domains for each channel following optimized filtering of the signals. The signals were split into planning and execution segments, respectively. Various delays and anticipation lengths were taken into account for each of the features, thus totaling 93 different features. The potential performance of each feature and channel for use in the classification of the EEG signals was analyzed by estimating the relative class overlap using the Davies-Bouldin Index (DBI), a widely used measure for estimating cluster separation. The best feature/channel configurations contained both channels that were close and channels that were far from motor areas. A statistical test using the channel/feature configurations that yielded the lowest 5% DBI values for motor and for non-motor channels yielded no significant difference (α= 0.05) between these two channel populations. The scope and depth of the study was limited. Plus, important parts of the signal had to be discarded to rule out interference stemming from saccadic eye movement However, our results do suggest more attention should be paid to non-motor areas in earlinked EEG data even when investigating movement related BCIs.
KW - brain-computer interface
KW - cluster separation
KW - EEG
KW - movement related potentials
KW - electroencephalography
KW - steady-state visual
KW - demodulation
KW - frequency domain analysis
KW - signal processing
KW - neural nets
KW - pattern recognition
UR - http://www.scopus.com/inward/record.url?scp=11244317428&partnerID=8YFLogxK
M3 - Conference contribution book
SN - 0780386442
SN - 9780780386440
SP - 942
EP - 946
BT - 2004 IEEE Conference on Cybernetics and Intelligent Systems
PB - IEEE
ER -