Non-linear Granger causality and its frequency decomposition in decoding human upper limb movement intentions. (1st January 2013)

• Record Type: Journal Article
• Authors: Liu, Meng Ting; Kuo, Ching-Chang; Chiu, Alan W.L.
• Is Part Of: International journal of biomedical engineering and technology. Volume 12:Number 1(2013)
• Page Start: 1
• Page End: 25
• Publication Date: 2013-01-01
• Subjects: Granger causality -- brain computer interface -- motor intention -- electroencephalogram; Biomedical engineering -- Periodicals; 610.28
• Languages: English
• ISSNs: 1752-6418

Abstract

Non-linear Granger Causality (GC) can detect directional influence of signals between multiple locations through a non-linear predictive approach using Radial Basis Functions (RBF). However, the directional relations in frequency domain are often lost. In this paper, we investigate how the nonlinear causal relationship at different frequency bands can be extracted by the proper linearisation process. The difference between the non-linearly and linearly fitted signals was used to determine if the frequency components can be approximated. This strategy was implemented and evaluated on a Brain Computer Interface (BCI) application to decode the intended directions of an imagined arm reaching movement (left, right and forward) using 128 surface electroencephalography (EEG) electrodes. The causal influence of the active recording sites was found to be unique with respect to the direction of the intended movement. Particularly, the left and right reaching directions were found to be highly separable in the theta rhythm (3–8 Hz).