Limits...
Temporal and spatial features of single-trial EEG for brain-computer interface.

Zhao Q, Zhang L - Comput Intell Neurosci (2007)

Bottom Line: Computer simulations on real experimental data indicate that independent residual analysis (IRA) method can provide efficient temporal features.Combining IRA features with the CSP method, we obtain the optimal spatial and temporal features with which we achieve the best classification rate.The high classification rate indicates that the proposed method is promising for an EEG-based brain-computer interface.

View Article: PubMed Central - PubMed

Affiliation: Department of Computer Science and Engineering, Shanghai Jiao Tong University, Shanghai 200240, China. qbzhao@sjtu.edu.cn

ABSTRACT
Brain-computer interface (BCI) systems create a novel communication channel from the brain to an output device bypassing conventional motor output pathways of nerves and muscles. Modern BCI technology is essentially based on techniques for the classification of single-trial brain signals. With respect to the topographic patterns of brain rhythm modulations, the common spatial patterns (CSPs) algorithm has been proven to be very useful to produce subject-specific and discriminative spatial filters; but it didn't consider temporal structures of event-related potentials which may be very important for single-trial EEG classification. In this paper, we propose a new framework of feature extraction for classification of hand movement imagery EEG. Computer simulations on real experimental data indicate that independent residual analysis (IRA) method can provide efficient temporal features. Combining IRA features with the CSP method, we obtain the optimal spatial and temporal features with which we achieve the best classification rate. The high classification rate indicates that the proposed method is promising for an EEG-based brain-computer interface.

No MeSH data available.


The classification accuracy versus the number offeatures for CSP, ICA and TSP (combination of IRA and CSP) methods. (a) SubjectA. (b) Subject B. (c) Subject C. (d) Subject D.
© Copyright Policy
Related In: Results  -  Collection

License
getmorefigures.php?uid=PMC2267213&req=5

fig13: The classification accuracy versus the number offeatures for CSP, ICA and TSP (combination of IRA and CSP) methods. (a) SubjectA. (b) Subject B. (c) Subject C. (d) Subject D.

Mentions: Table 1 summarizes the results of single-trial EEGclassification for left- versus right-hand movement imagery. The first rowdenotes the different classification method with different number of features,the first column denotes different feature extraction methods for the subjects.In the feature extraction methods, temporal spatial pattern (TSP) representsthe method of combining IRA and CSP which we have proposed in this paper. Inthe table, ICA results are computed by infomax ICA technique throughdecomposing the data into 62 components and then selecting different number offeatures based on mutual information method. From the table, we can see thatCSP algorithm is sensitive for the frequency (i.e., frequency-specific). ICAresults have no obvious improvement with increasing number of features. We alsosee clearly that the TSP method improves the accuracy of classification.Without applying filtering on EEG signals, TSP method always get better resultsthan the CSP algorithm. Furthermore, Figure 13 shows the curves ofclassification rate according to number of features. The most optimal resultcan be obtained by the TSP method and the accuracy is about 93.9% for subject A,95% for subject B, 92.33% for subject C, and 91.3% for subject D. In the GrazBCI system, subjects were asked to perform the actual finger movement at 8second and the system also has the feedback to subjects at 1 second after themovement according to the estimate of DSLVQ classifier. However, in our system,the subject only was asked to imagine hand movement but none of actual movementand feedback were performed. In fact, the actual movement will improve theclassification rate greatly. Moreover, there is no preselection for artifacttrials in our system. Therefore, TSP can provide better features for EEGclassification during hand movement imagery and is more suitable for the onlineBCI system.


Temporal and spatial features of single-trial EEG for brain-computer interface.

Zhao Q, Zhang L - Comput Intell Neurosci (2007)

The classification accuracy versus the number offeatures for CSP, ICA and TSP (combination of IRA and CSP) methods. (a) SubjectA. (b) Subject B. (c) Subject C. (d) Subject D.
© Copyright Policy
Related In: Results  -  Collection

License
Show All Figures
getmorefigures.php?uid=PMC2267213&req=5

fig13: The classification accuracy versus the number offeatures for CSP, ICA and TSP (combination of IRA and CSP) methods. (a) SubjectA. (b) Subject B. (c) Subject C. (d) Subject D.
Mentions: Table 1 summarizes the results of single-trial EEGclassification for left- versus right-hand movement imagery. The first rowdenotes the different classification method with different number of features,the first column denotes different feature extraction methods for the subjects.In the feature extraction methods, temporal spatial pattern (TSP) representsthe method of combining IRA and CSP which we have proposed in this paper. Inthe table, ICA results are computed by infomax ICA technique throughdecomposing the data into 62 components and then selecting different number offeatures based on mutual information method. From the table, we can see thatCSP algorithm is sensitive for the frequency (i.e., frequency-specific). ICAresults have no obvious improvement with increasing number of features. We alsosee clearly that the TSP method improves the accuracy of classification.Without applying filtering on EEG signals, TSP method always get better resultsthan the CSP algorithm. Furthermore, Figure 13 shows the curves ofclassification rate according to number of features. The most optimal resultcan be obtained by the TSP method and the accuracy is about 93.9% for subject A,95% for subject B, 92.33% for subject C, and 91.3% for subject D. In the GrazBCI system, subjects were asked to perform the actual finger movement at 8second and the system also has the feedback to subjects at 1 second after themovement according to the estimate of DSLVQ classifier. However, in our system,the subject only was asked to imagine hand movement but none of actual movementand feedback were performed. In fact, the actual movement will improve theclassification rate greatly. Moreover, there is no preselection for artifacttrials in our system. Therefore, TSP can provide better features for EEGclassification during hand movement imagery and is more suitable for the onlineBCI system.

Bottom Line: Computer simulations on real experimental data indicate that independent residual analysis (IRA) method can provide efficient temporal features.Combining IRA features with the CSP method, we obtain the optimal spatial and temporal features with which we achieve the best classification rate.The high classification rate indicates that the proposed method is promising for an EEG-based brain-computer interface.

View Article: PubMed Central - PubMed

Affiliation: Department of Computer Science and Engineering, Shanghai Jiao Tong University, Shanghai 200240, China. qbzhao@sjtu.edu.cn

ABSTRACT
Brain-computer interface (BCI) systems create a novel communication channel from the brain to an output device bypassing conventional motor output pathways of nerves and muscles. Modern BCI technology is essentially based on techniques for the classification of single-trial brain signals. With respect to the topographic patterns of brain rhythm modulations, the common spatial patterns (CSPs) algorithm has been proven to be very useful to produce subject-specific and discriminative spatial filters; but it didn't consider temporal structures of event-related potentials which may be very important for single-trial EEG classification. In this paper, we propose a new framework of feature extraction for classification of hand movement imagery EEG. Computer simulations on real experimental data indicate that independent residual analysis (IRA) method can provide efficient temporal features. Combining IRA features with the CSP method, we obtain the optimal spatial and temporal features with which we achieve the best classification rate. The high classification rate indicates that the proposed method is promising for an EEG-based brain-computer interface.

No MeSH data available.