Study On Non-contact System For Drive Fatigue Detection Based On Multi-source Information Fusion

With the improvement of living standard in modern societies,car ownership is significantly increasing yearly,which led to frequent occurrence of traffic accident.Drive's fatigue is the most common causes for road accidents.Current techniques for drive's fatigue detection are intrusive as electrode attachment to the driver is required.This usually proves to be as annoyance to the driver and is in most situation interfering driver's operation.Non-contact image sensor system to measure facial expression of driver,although it overcomes the problem of contact detection,accuracy and reliability are low.In addition,the technique tends to be sensitive to external factors such as luminance or driver appearance.Therefore,the study of a multi-source information fusion as a non-contact fatigue detection system can effectively reduce traffic accidents.In this paper,a multi-source information fusion system based on detections of physiological signals and steering angle change is designed and applied for driver's fatigue's detection.The system includes a Doppler radar for physiological signal detection and steering wheel angle detection system.A video recording system is used as a standard for drive's fatigue.Firstly,based on the principle of Doppler radar detection of physiological signals,the hardware system obtained physiological information through radar signal preprocessing,amplification,active band-pass filtering,analog-digital conversion and a series of signal processing.Steering angle acquisition system by combining Hall angle sensor and rotary encoder approaches were used to obtain the steering wheel angle change information.Secondly,based on the characteristics of radar digital signal,two different digital filtering methods of FIR and IIR were used to separate the physiological signals.The advantages and disadvantages of the algorithm are compared.The zero-phase IIR filter algorithm is chosen to extract the respiratory and heartbeat signals.Finally,the statistical analysis by student T test was performed on the obtained physiological signal data after comparing with facial expression change recorded by video camera.The values of physiological signals and steering wheel signals were extracted under different degrees of fatigue,trained by the extreme learning machine algorithm and the fatigue sample data were identified and tested.Conclusion: We found that decreases in respiratory wave levels are strongly correlated with drive fatigue.The waveforms of the experimental data show that the steering wheel angle changes,the driver's breathing rate and respiratory rate are reduced with the deepening of sleep.The recognition test of the driver fatigue samples was carried out,and the results showed that the recognition rate was 81% in the fatigue state.Therefore,multi-source information fusion system canbe used as a non-contact device for fatigue detection during driving with predictive and high recognition accuracy.