Multi-domain Feature of Event-Related Potential Extracted by Nonnegative Tensor Factorization: 5 vs. 14 Electrodes EEG Data

Abstract

Abstract. As nonnegative tensor factorization (NTF) is particularly useful for the problem of underdetermined linear transform model, we performed NTF on the EEG data recorded from 14 electrodes to extract the multi-domain feature of N170 which is a visual event-related potential (ERP), as well as 5 typical electrodes in occipital-temporal sites for N170 and in frontal-central sites for vertex positive potential (VPP) which is the counterpart of N170, respectively. We found that the multi-domain feature of N170 from 5 electrodes was very similar to that from 14 electrodes and more discriminative for different groups of participants than that of VPP from 5 electrodes. Hence, we conclude that when the data of typical electrodes for an ERP are decomposed by NTF, the estimated multi-domain feature of this ERP keeps identical to its counterpart extracted from the data of all electrodes used in one ERP experiment.

A synchronous and multi-domain feature extraction method of EEG and sEMG in power-assist rehabilitation robot

“…Recently, the NTF has been used in the study of bio-informatics, image understanding and neuroscience. NTF has been used in EEG analysis about cognitive analysis and motor imagery [18,19], whereas, NTF in synchronous and multi-domain feature extraction of EEG and sEMG in neural control is less intensively studied.…”

Sequential nonnegative tucker decomposition on multi-way array of time-frequency transformed event-related potentials

“…Thus, fast tensor factorization algorithms are desired. In this study, we examine nonnegative Tucker decomposition (NTD) algorithms to extract the multi-domain features [7] of ERPs.…”

“…However, practically, the randomly fluctuating part cannot be sufficiently rejected and the constant part may include overlapped components. Consequently, except conventional signal processing methods like optimal digital filter and wavelet filter [4,5], group independent component analysis (ICA) [6] and nonnegative tensor factorization [7] have been used for the group-level analysis of ERPs. With the advanced methods, artifacts can be further rejected and desired components can be extracted simultaneously [8,9].…”

“…Regarding such applications, a third-order tensor usually includes timefrequency representation (TFR) of EEG at multiple channels [10][11][12][13][14], and the fourth-order one often consists of TFR of EEG of multiple subjects at multiple channels [7,20]. Particularly, for the study of ERPs, facilitating the grouplevel analysis requires decomposing the tensor including data of different groups of participants in different conditions together [7,20]. Hence, the sizes of decomposed tensor are immense, which makes tensor factorization computationally inefficient.…”

“…Tensor factorization has been applied to study spontaneous EEG [10][11][12][13][14][15], single-trial EEG in an ERP experiment [16][17][18], and ERPs [7,[19][20][21], for brain computer interface and for the cognitive research. Regarding such applications, a third-order tensor usually includes timefrequency representation (TFR) of EEG at multiple channels [10][11][12][13][14], and the fourth-order one often consists of TFR of EEG of multiple subjects at multiple channels [7,20].…”