Automatic design for independent component analysis based brain-computer interfacing

Publication Type:
Conference Proceeding
Proceedings of the Annual International Conference of the IEEE Engineering in Medicine and Biology Society, EMBS, 2013, pp. 2180 - 2183
Issue Date:
Filename Description Size
06609967.pdfPublished version1.26 MB
Adobe PDF
Full metadata record
This study proposes a new framework, independent component ensemble, to leverage the acquired knowledge into a truly automatic and on-line EEG-based brain-computer interfacing (BCI). The envisioned design includes: (1) independent source recover using independent component analysis (ICA) (2) automatic selection of the independent components of interest (ICi) associated with human behaviors; (3) multiple classifiers with a parallel constructing and processing structure; and (4) a simple fusion scheme to combine the decisions from multiple classifiers. Its implications in BCI are demonstrated through a sample application: cognitive-state monitoring of participants performing a realistic sustained-attention driving task. Empirical results showed the proposed ensemble design could provide an improvement of 7%∼15% in overall accuracy for the classification of the arousal state and the driving performance. In summary, constructing ICi-ensemble classifiers and combining their outputs demonstrates a practical option for ICA-based BCIs to reduce the risk of not obtaining any desired independent source or selecting an inadequate component. Most importantly, the ensemble design for integrating information across multiple brain areas creates potentials for developing more complicated BCIs for real world applications. © 2013 IEEE.
Please use this identifier to cite or link to this item: