Research Of The Mental Fatigue Classification Based On Physiological Electrical Signals

Intense social competition and rapid pace of life lead to serious psychological pressure and more and more people who feel fatigue. Mental fatigue cause the lower efficiency of work or study,and constantly casuse various physical and mental illness,serious threat to human health. Meanwhile mental fatigue is also important factor resulting in transportation,production and other accidents. Therefore, research of mental fatigue have not only important medical value, but also the great social value and economic value.Physiological signals include a large number of human physiological information, can effectively reflect the changes in physiological state, has become a reliable indicator of fatigue state assessment. Recently, biomedical signal processing technology is developing rapidly, the development of digital signal processing and pattern recognition methods lead to the great progress of the research of recognising mental fatigue state.To explore the relationship between metal fatigue and the physiological signals, this stduy designs a mental fatigue experiment with two hours mental arithmetic. We collect data, including VAS, mental arithmetic performance and Physiological signals(EEG, EOG, etc).In this stduy, it lead to the analysis of EEG become difficult, that EEG is mixing of Ocular Artifacts. For the Ocular Artifacts which frequent appeared on EEG, a method of ocular artifacts removal was proposed, which is based on the threshold thinking and empirical mode decomposition method, and the experiment results show the effectiveness of the method.By the statistical analysis of VAS and mental arithmetic performance, and physiological signals analysis in time domain, frequency domain, nonlinear analysis, three kinds of fatigue level (mild fatigue, moderate fatigue, severe fatigue) is selected,8 characteristics, like EMG standard deviation, which reflect mental fatigue information is proposed. Finally, the fatigue level is identified by support vector machine. When the SVM training set and test set from the same subjects, the recognitiong rate of three fatigue level reach 95%. And the recognition rate is 66.7% caused by individual differences in physiological signal recognition, when training, test from some different subjects. The classification results prove the validity of proposed feature vector, and physiological signals can reflect fatigue level.