Signal Processing Methods In Driver Fatigue And Sub-health Research

Along with the social progress and development, the problems caused by fatigue have emerged, affecting seriously the people's normal study, work and life. Especially the harmfulness of fatigue driving and sub-health are enormous, they have been one of the "killer" of people's life. Therefore, the scientific, rational, objectively detect fatigue driving and sub-health and provide interference basis for relevant department have very important practical significance.Physiological signal contains massive information which closely related human body's physiological activity, mental state and disease diagnosis. Therefore, detection of fatigue driving and sub-health can be achieved by analysis physiological signal. The analysis of the body's physiological signal is one of the typical applications of signal processing technology. It can effectively extract the intrinsic characteristics and essential information of the physiological signal. Based on the existed theories, this article uses the modern signal processing technology to study fatigue driving and sub-health from the point of view of physiological signal. And it specific collects the EEG signals from12subjects in the process of a driving simulation and the pulse signals from30subjects to detect fatigue driving and sub-health state, and then explore the effective characteristics of driver fatigue and sub-health state. The main work of this study is:Firstly, EEG signals and pulse signals are de-noised respectively through Hilbert-Huang transform and Wavelet transform, in order to improve the accuracy of the analysis. By comparison with signal waveform before de-noising and signal waveform after de-noising, found that Hilbert-Huang transform and Wavelet transform can effectively remove the interference signal which be mingled with the EEG signal and pulse signal, so as to lay a foundation for further signal analysis.Secondly, relative power spectrum, Wigner-Ville distribution and power spectrum information entropy are respectively used to analyze the EEG signals at different driving times, finding that EEG rhythms' relative power spectrum and spectrum information entropy values have existed obvious differences at different driving times, and the Wigner-Ville distributions of EEG are also not the same at different times, which demonstrates the relative power spectrum, Wigner-Ville distribution and power spectrum information entropy can be used for detecting fatigue driving.Finally, Matching pursuit algorithm and Gabor transform are adopted to extract pulse signals' features of the health and sub-health, and have carried on the analysis and comparison, then the health and sub-health state are classified by K-nearest neighbors classifier based on the selected feature as the input vector, and achieves good classification effect, which indicates that the Matching pursuit algorithm and Gabor transform can be used for detecting sub-health state.In sum, the relative power spectrum, Wigner-Ville distribution and power spectrum information entropy can be used to analyze EEG feature in the process of driving, with a result whether a driver is at fatigue state. The Matching pursuit algorithm and Gabor transform can be used to extract the pulse signal features of sub-health and can obtain good classification effect, thus they might be considered as reference index of sub-health detection. Thus, it can be seen that signal processing methods can effectively extract the characterization of fatigue driving and sub-health state, and open up a practical way for the driver fatigue detection and diagnosis of sub-health.