A Probabilistic Framework for Learning Robust Common Spatial Patterns

Department of Brain and Cognitive Sciences, Massachusetts Institute of Technology, Cambridge, MA 02139, USA.
Conference proceedings: ... Annual International Conference of the IEEE Engineering in Medicine and Biology Society. IEEE Engineering in Medicine and Biology Society. Conference 09/2009; 2009:4658-61. DOI: 10.1109/IEMBS.2009.5332646
Source: IEEE Xplore


Robustness in signal processing is crucial for the purpose of reliably interpreting physiological features from noisy data in biomedical applications. We present a robust algorithm based on the reformulation of a well-known spatial filtering and feature extraction algorithm named Common Spatial Patterns (CSP). We cast the problem of learning CSP into a probabilistic framework, which allows us to gain insights into the algorithm. To address the overfitting problem inherent in CSP, we propose an expectation-maximization (EM) algorithm for learning robust CSP using from a Student-t distribution. The efficacy of the proposed robust algorithm is validated with both simulated and real EEG data.

Download full-text


Available from: Xiaorong Gao, Jul 18, 2014
  • Source
    • "Alternatively, the non-stationarity issue has also been addressed within a regularization framework in [32], [33], and via a cluster-based approach in [34]. Finally, several robust CSP algorithms have been developed to alleviate the sensitivity to noise and outliers [35], [36], [37], [38] "
    [Show abstract] [Hide abstract]
    ABSTRACT: Common spatial patterns (CSP) is a well-known spatial filtering algorithm for multichannel electroencephalogram (EEG) analysis. In this paper, we cast the CSP algorithm in a probabilistic modeling setting. Specifically, probabilistic CSP (P-CSP) is proposed as a generic EEG spatio-temporal modeling framework that subsumes the CSP and regularized CSP algorithms. The proposed framework enables us to resolve the overfitting issue of CSP in a principled manner. We derive statistical inference algorithms that can alleviate the issue of local optima. In particular, an efficient algorithm based on eigendecomposition is developed for maximum a posteriori (MAP) estimation in the case of isotropic noise. For more general cases, a variational algorithm is developed for group-wise sparse Bayesian learning for the P-CSP model and for automatically determining the model size. The two proposed algorithms are validated on a simulated data set. Their practical efficacy is also demonstrated by successful applications to single-trial classifications of three motor imagery EEG data sets and by the spatio-temporal pattern analysis of one EEG data set recorded in a Stroop color naming task.
    Full-text · Article · Jan 2015 · IEEE Transactions on Pattern Analysis and Machine Intelligence
  • Source
    • ". Graphical representations of the PCSP [11] and BCSP models [6]. "
    [Show abstract] [Hide abstract]
    ABSTRACT: Multi-subject electroencephalography (EEG) classification involves the categorization of brain waves measured from multiple subjects, each of whom undergoes the same mental task. Common spatial patterns (CSP) or probabilistic CSP (PCSP) are widely used for extracting discriminative features from EEG, although they are trained on a subject-by-subject basis and inter-subject information is neglected. Moreover, the performance is degraded when only a few training samples are available for each subject. In this paper, we present a method for Bayesian CSP with Dirichlet process (DP) priors, where spatial patterns (corresponding to basis vectors) are simultaneously learned and clustered across subjects using variational Bayesian inference, which facilitates a flexible mixture model where the number of components are also learned. Spatial patterns in the same cluster share the hyperparameters of their prior distributions, which means information transfer is facilitated among subjects with similar spatial patterns. Numerical experiments using the BCI competition IV 2a dataset demonstrated the high performance of our method, compared with existing PCSP and Bayesian CSP methods with a single prior distribution.
    Preview · Conference Paper · Jun 2012
  • Source
    • "Here we briefly explain about CSP in the context of EEG signal processing [2] [3]. Let us assume the observation data as X ∈ R N ×T , where N is the number of channels (electrode leads) and T is the number of time samples. "
    [Show abstract] [Hide abstract]
    ABSTRACT: In this paper, we aim to identify the regions involved in epilepsy from intracerebral EEG (iEEG) of patients suffering from focal epilepsy. Identification of regions involved in epilepsy is important for presurgery evaluations. The proposed method is based on common spatial pattern (CSP) using two types of time intervals: 1) periods including interictal epileptiform discharges (IED), and 2) periods excluding IEDs or abnormal physiological signals. The method is applied on the iEEG recordings of one seizure-free patient after resective surgery. The results are compared with seizure onset zones visually inspected by the epileptologist. The congruent IED regions with visually detected seizure onset zones are encouraging results. Moreover, the application of CSP method for the identification of IED regions seems interesting as this method is fast and simple.
    Full-text · Article · Jun 2011
Show more