Research On Brain Fatigue Detection Method Based On Deep Learning

With the rapid development of science,technology,and economic levels,people’s lives are becoming increasingly fast-paced,and they are facing growing competition pressure.Prolonged engagement in intensive mental work can lead to mental fatigue.In daily commuting,automobiles play an increasingly important role.Driving is a form of intense mental labor,and drivers experience fatigue after continuous driving for more than 4 hours,leading to fatigue driving.In certain specialized fields such as mining,clinical,and high-altitude operations,fatigue can also result in production accidents causing unforeseen harm.Brain fatigue can manifest as fatigue,reduced cognitive abilities,and difficulty concentrating.Prolonged periods of mental fatigue can also lead to physiological ailments such as heart disease and palpitations,as well as psychological disorders such as insomnia,anxiety,and depression,severely affecting physical and mental health.Therefore,if a brain fatigue identification model can be explored and developed,it can real-time recognize the fatigue state of operators and provide warnings,allowing them to take timely breaks or use other relaxation measures,thereby avoiding fatigue-related accidents and protecting people’s physical and mental well-being.The main research of this paper is as follows:(1)The 2-back brain fatigue induction experiment is proposed as the experimental paradigm,and brainwave data of 10 healthy adult volunteers are collected to establish a brain fatigue dataset.During the experiment,subjective fatigue scores from a fatigue rating scale are used as the evaluation index for the fatigue state of the subjects.Based on the scale scores,the brain fatigue state is divided into three categories:alert state,mild fatigue state,and severe fatigue state.(2)Preprocessing of the raw EEG data is conducted before using neural networks for detection.Independent component analysis is used to remove artifacts,and then discrete wavelet transform is applied to extract EEG time-frequency features.Subsequently,convolutional neural network models and convolutional neural network combined with long short-term memory network models are designed for fatigue detection.Based on performance comparison results,the convolutional neural network combined with the long short-term memory network is ultimately selected as the fatigue detection model,achieving an average recognition accuracy of 97.12%,sensitivity of 97.80%,and specificity of 99.28%,proving the effectiveness of this model for fatigue detection tasks.(3)By using power spectral density as a feature to generate brain topography based on EEG rhythms,the study further reveals that the changes in subjects’ brainwaves during the experiment are mainly concentrated in the frontal lobe area,which is verified through channel selection experiments.The experimental results demonstrate the ability of this research to accurately distinguish between alertness,mild brain fatigue,and severe brain fatigue states.It also elucidates the relationship between changes in volunteers’ fatigue states and EEG channels during fatigue induction experiments,contributing to the further application of deep learning models in the field of brain fatigue detection.