Nonnegative Tensor Factorization for Continuous EEG Classification

Department of Computer Science, Pohang University of Science and Technology, San 31 Hyoja-dong, Nam-gu, Pohang 790-784, Republic of Korea.
International Journal of Neural Systems (Impact Factor: 6.51). 09/2007; 17(4):305-17. DOI: 10.1142/S0129065707001159
Source: PubMed


In this paper we present a method for continuous EEG classification, where we employ nonnegative tensor factorization (NTF) to determine discriminative spectral features and use the Viterbi algorithm to continuously classify multiple mental tasks. This is an extension of our previous work on the use of nonnegative matrix factorization (NMF) for EEG classification. Numerical experiments with two data sets in BCI competition, confirm the useful behavior of the method for continuous EEG classification.

Download full-text


Available from: Lee Hyekyoung, Jun 24, 2014
  • Source
    • "To demonstrate the performance of the proposed method, we compared it with two state-of-the-arts methods: (1) CPD for feature extraction and support vector machine (SVM) [10] for classification (2) common spatial patterns (CSP) [11], [12] for feature extraction and SVM for classification. The former method is generally used for CPDbased classification [2]. The latter method is a classical method used in the brain computer interface, which has a good performance, especially for motor imagery EEG. "
    [Show abstract] [Hide abstract]
    ABSTRACT: Canonical Polyadic Decomposition is an important method for multi-way array analysis, and has been successfully applied to a variety of signals. It can be seen as an unsupervised method for feature extraction in a classification problem. After that, a classifier, such as support vector machine, is utilized to classify those features. In this manner, classification task is achieved by two isolated steps. We proposed supervised Canonical Polyadic Decomposition by directly incorporating auxiliary label information during decomposition, with which a classification task can be achieved without an extra step of classifier training. The proposed method integrates decomposition and classifier learning together, so it reduces procedure of classification task compared with that of separated decomposition and classification. In order to evaluate the performance of the proposed method, three different kinds of signals, synthetic signal, EEG signal, and MEG signal, are used. The results based on evaluations of synthetic and real signals demonstrated the proposed method is effective and efficient.
    Full-text · Article · Oct 2014
  • Source
    • "In the recent years, a lot of attention has been paid to blind source separation (BSS) due to its wide-ranging applications in many areas [1] such as audio and speech processing [2], telecommunications [3], biomedical engineering [4], and hyperspectral imaging [5]. Assuming an M-dimensional observation vector, yðkÞ, this problem is mathematically expressed as yðkÞ ¼ AxðkÞþbðkÞ; ð1Þ where xðkÞ denotes the N-dimensional source vector, bðkÞ denotes the M-dimensional additive noise vector, and A is the M Â N mixing matrix. "
    [Show abstract] [Hide abstract]
    ABSTRACT: This paper addresses the problem of blind source separation for underdetermined mixtures (i.e., more sources than sensors) of event-related sources that include quasi-periodic sources (e.g., electrocardiogram (ECG)), sources with synchronized trials (e.g., event-related potentials (ERP)), and amplitude-variant sources. The proposed method is based on two steps: (i) tensor decomposition for underdetermined source separation and (ii) signal extraction by Kalman filtering to recover the source dynamics. A tensor is constructed for each source by synchronizing on the “event” period of the corresponding signal and stacking different periods along the second dimension of the tensor. To cope with the interference from other sources that impede on the extraction of weak signals, two robust tensor decomposition methods are proposed and compared. Then, the state parameters used within a nonlinear dynamic model for the extraction of event-related sources from noisy mixtures are estimated from the loading matrices provided by the first step. The influence of different parameters on the robustness to outliers of the proposed method is examined by numerical simulations. Applied to clinical electroencephalogram (EEG), ECG and magnetocardiogram (MCG), the proposed method exhibits a significantly higher performance in terms of expected signal shape than classical source separation methods such as π CA and FastICA.
    Full-text · Article · Aug 2014 · Signal Processing
  • Source
    • "To our knowledge, EEG artifact removal within a joint processing framework of single-channel EEG and auxiliary recordings is completely novel. Also, while NTF has already been used for EEG-feature extraction [7], its use for EEG artifact removal within a Gaussian source separation framework is novel. "
    [Show abstract] [Hide abstract]
    ABSTRACT: New applications of Electroencephalographicrecording (EEG) require light and easy-to-handle equipment involving powerful algorithms of artifact removal. In our work, we exploit informed source separation methods for artifact removal in EEG recordings with a low number of sensors, especially in the extreme case of single-channel recording, by exploiting prior knowledge from auxiliary lightweight sensors capturing artifactual signals. To achieve this, we propose a method using Non-negative Tensor Factorization (NTF) in a Gaussian source separation framework that proves competitive against the classic Independent Component Analysis (ICA) technique. Additionally the both NTF and ICA methods are used in an original scheme that jointly processes the EEG and auxiliary signals. The adopted NTF strategy is shown to improve the source estimates accuracy in comparison with the usual multi-channel ICA approach.
    Full-text · Conference Paper · Sep 2013
Show more