Research On Operator Fatigue Model Based On ECG Signal

With the rapid development of automation technology,the working ways of human operators have changed enormously,gradually shifting from manual operation to monitoring the operating status of automatic control systems.The research shows that fatigue operation is one of the important reasons leading to accidents.As a result,the research on the fatigue condition of operators has concerned widely around the world.For the purpose of reducing the probability of accidents caused by operating the automatic control system in the fatigue state of the operator,and ensuring the safe operation of the operator,it is very significant to detect the fatigue state of operator immediately and accurately.Although the traditional ECG fatigue classification method can effectively identify the fatigue state,it needs to collect a long time signal and cannot achieve the real-time monitoring of the fatigue state.Compared with traditional algorithms,the deep learning method has the advantage that it does not require a complicated feature extraction process,and can detect fatigue status in real time through short-term ECG signals,thereby reducing accidents.In this paper,the following researches are carried out on the fatigue evaluation of operators:(1)The current research on the fatigue state of operators is mainly applied in aviation,automobile driving,rail driving and other fields.In this paper,fatigue experiments are designed based on the Automation-enhanced Cabin Air Management System.During the experiment,the EEG1100 physiological signal acquisition device was used to collect the ECG signals of 8 subjects under different task difficulties,which provided a data set for the subsequent establishment of the fatigue model.In addition,the visual analog scale method was used to evaluate the subjective fatigue of the subjects.(2)A one-dimensional double convolutional neural network(1D-ECNN)method is proposed to detect the fatigue state of the operator based on the collected ECG signals.The1D-ECNN network includes 2 convolutional blocks,1 fully connected layer,and 1 softmax output layer.Each convolutional block is composed of two convolutional layers and a maximum pooling layer.This method uses only a small number of convolution kernels,which can reduce the intricacies of model and increase the speed of model training.Meanwhile,it avoids the complicated feature extraction process or feature selection process in traditional methods.The ECG signals are divided into samples with a time length of 1s,and a one-dimensional dual convolution neural network fatigue classification model is established based on the short-term ECG signal.The emulation result shows that the average sort accuracy of the approach reaches up to 95.72%,which and can detect the fatigue state of the operator timely and exactly.(3)Under the condition of ensuring accuracy,in order to solve the problem of less ECG data in this article,a deep convolutional neural network(DCNN)based on transfer learning is designed to evaluate the fatigue state of the operator and to realize the automatic classification of the short-term ECG signal fatigue state of the operator.First,a method of converting the ECG signal into an image is proposed,which transforms the collected ECG signal into a 2D image,that is,the ECG signal is directly mapped to the 2D space and converted into time-domain picture information.Then,the pictures are sent to the DCNN model for training to realize the classification of operator fatigue.The consequences indicate that the method can extract effective features from the ECG signal automatically.Compared with 1D-ECNN,it achieves a better fatigue classification effectand realizes the correct classification of the non-fatigue and fatigue states of the operator.The average classification accuracy rate reaches 97.36%.At the same time,it can better eliminate the influence of individual differences.