Research On Non-contact Fatigue Driving Detection Method Based On Regularized Extreme Learning Machine

Non-contact fatigue driving detection based on neural network has become a hot research direction in the field of fatigue driving detection.It effectively solves the problem that the contact fatigue detection method brings to the driver and the low reliability of the single signal source to reflect the fatigue degree.At the same time,the multi-source information is classified by designing the neural network model to achieve high precision of the fatigue state.And high speed detection.Choosing the appropriate eigenvalues is critical to the accuracy of network detection and to accurately reflect fatigue.Reliability and accuracy are high based on driver physiological signal detection.However,collecting physiological signals by attaching electrodes to the body still causes interference to the driver.In addition,physiological signals have different physiological signals in different fatigue states and different physiological signals in different individuals under different fatigue states.The accuracy of classification can be reduced by directly using the characteristic values of the original physiological signals as network input.Therefore,it is necessary to preprocess the eigenvalues.In view of the above problems,this paper proposes a non-contact method for detecting the fatigue state of the driver based on the extreme learning machine,which can effectively improve the detection accuracy of the fatigue state.The method mainly includes three parts: data acquisition,data processing and data training.The data acquisition part uses the Doppler radar module to collect the physiological signals of the driver;the data processing part uses the expert evaluation method to classify the data,and performs preprocessing and feature extraction to ensure the data transformation of the data within the controllable range of the error.Effectively solve individual differences.The Regularized Extreme Learning Machine(RELM)is designed to train the data set to obtain an algorithm model for driver fatigue state detection.In this paper,the driver’s physiological signal is collected by Doppler radar module,and the driver’s physiological signal original data set is obtained through post-processing.It is named as the DOPS(Driver’s Original Physiological Signal)data set,and the driver’s physiological signal mapping data is further obtained through feature transformation.The set is named DPSR(Driver Physiological Signal Ratio)data set.In this paper,we first design the RELM algorithm model,and carry out experiments on the DOPS data set and the DPSR data set respectively.The experimental results show that the prediction accuracy of the fatigue state using the DPSR data set is improved by 17% compared with the DOPS data set,which proves that the eigenvalue is performed.The effectiveness of the transformation process.In order to verify the superiority of the RELM algorithm in terms of fatigue state classification and high detection accuracy,the author also designed a momentum-optimized BP(Back Propagation)neural network algorithm model to perform experiments on the DPSR data set.The detection result RELM improves the training speed.The accuracy of the detection is higher.