Abstract

Info

Author(s):: LIU Jiazhuo; XIE Yun; CHEN Xueqiang; WU Yang; School of Automation, Guangdong University of Technology, Guangzhou 510006, China

Keywords:: ; Keywords: steady-state visual evoked potential; Hilbert-Huang algorithm; visual fatigue characteristic; support vector machine

Abstract:: Abstract: An electroencephalogram (EEG) experiment based on steady-state visual evoked potentials is designed to accurately extract the fatigue characteristics of EEG signals and use those characteristics as warning signs to remind the programmer to rest. The Hilbert marginal spectral energy distributions of EEG parameters, namely[θ],[α],[β],[βα]and [α+θβ], in the normal and the fatigue states are extracted by Hilbert-Huang algorithm. The change trends of Hilbert marginal spectrum energy in both states are analyzed. The results of one-way analysis of variance reveal that in the fatigue state, the marginal spectral energy of [α] and [α+θβ]increases significantly, and that the marginal spectral energy of [β]and [βα]decreases significantly. By support vector machine classification, the maximum classification accuracy rates of [βα] and [β] reach 94.4% and 93.3%, respectively. A good separability is also found between[α]and[α+θβ]. The marginal spectral energy characteristics of the 4 EEG parameters, namely[α],[β],[βα],[α+θβ], extracted with Hilbert-Huang algorithm can be used as indicators to evaluate the visual fatigue.