2009 4th International IEEE/EMBS Conference on Neural Engineering 2009
DOI: 10.1109/ner.2009.5109293
|View full text |Cite
|
Sign up to set email alerts
|

Subclass discriminant analysis using dynamic cluster formation for EEG-based brain-computer interface

Help me understand this report

Search citation statements

Order By: Relevance

Paper Sections

Select...
2
1

Citation Types

0
3
0

Year Published

2011
2011
2022
2022

Publication Types

Select...
5

Relationship

0
5

Authors

Journals

citations
Cited by 5 publications
(3 citation statements)
references
References 8 publications
0
3
0
Order By: Relevance
“…2.1) Subclass Divisions: How to divide each class into different subclasses is a crucial problem in SDA. Different unsupervised algorithms have been attempted in this field, including K-means clustering [25], [28], dynamic cluster formation [26], Gaussian mixture model [29], hierarchical clustering [24], nearest neighbor (NN) clustering [22], and valley seeking algorithm [27]. However, these unsupervised methods usually clustered data in terms of their inherent similarity, overlooking the physical meanings of the clustered data.…”
Section: Data Processing and Analysismentioning
confidence: 99%
“…2.1) Subclass Divisions: How to divide each class into different subclasses is a crucial problem in SDA. Different unsupervised algorithms have been attempted in this field, including K-means clustering [25], [28], dynamic cluster formation [26], Gaussian mixture model [29], hierarchical clustering [24], nearest neighbor (NN) clustering [22], and valley seeking algorithm [27]. However, these unsupervised methods usually clustered data in terms of their inherent similarity, overlooking the physical meanings of the clustered data.…”
Section: Data Processing and Analysismentioning
confidence: 99%
“…1). The research on SDA usually emphasize the separability of original classes rather than subclasses, ignoring the potential meaning of subclasses [25], [26], [27], [28]. Our previous study showed that the subclass information of each finger gesture can be used to represent the wrist rotation position [22].…”
Section: Machine Learning Algorithmsmentioning
confidence: 99%
“…Therefore, the subspace that maximizes the constrained log-likelihood function in (19) at each EM cycle coincides with the subspace that maximizes the MSDA criterion, where the scatter matrices in (4) are replaced by their weighted equivalent in each EM cycle. The MLE of the true means can now be computed by substituting (23), (32) into (11) and using the computed estimates of ( 16), (35) for S ws and Ψ respectively…”
Section: Em Algorithmmentioning
confidence: 99%