L1-Norm-Based Common Spatial Patterns

• Published in: IEEE transactions on biomedical engineering 2012-03, Vol.59 (3), p.653-662
• Main Authors: Wang, Haixian, Tang, Qin, Zheng, Wenming
• Format: Article
• Language: English
• Subjects: Algorithms; Applied sciences; Artificial Intelligence; Biological and medical sciences; Brain Mapping; Brain-computer interfaces; common spatial patterns (CSP); Computer science; control theory; systems; Computer Simulation; Computer systems and distributed systems. User interface; Computerized, statistical medical data processing and models in biomedicine; CSP-L1; Dispersion; Eigenvalues and eigenfunctions; Electroencephalography; Electroencephalography - methods; Exact sciences and technology; Humans; Information, signal and communications theory; Integrated circuits; Iterative methods; L1-norm; Medical sciences; Models and simulation; Models, Theoretical; Pattern recognition; Pattern Recognition, Automated; Reproducibility of Results; robust modeling; Robustness; Signal processing; Signal Processing, Computer-Assisted; Software; Telecommunications and information theory; Vectors
• ISSN: 0018-9294; 1558-2531
• DOI: 10.1109/TBME.2011.2177523

Common spatial patterns (CSP) is a commonly used method of spatial filtering for multichannel electroencephalogram (EEG) signals. The formulation of the CSP criterion is based on variance using L2-norm, which implies that CSP is sensitive to outliers. In this paper, we propose a robust version of CSP, called CSP-L1, by maximizing the ratio of filtered dispersion of one class to the other class, both of which are formulated by using L1-norm rather than L2-norm. The spatial filters of CSP-L1 are obtained by introducing an iterative algorithm, which is easy to implement and is theoretically justified. CSP-L1 is robust to outliers. Experiment results on a toy example and datasets of BCI competitions demonstrate the efficacy of the proposed method.