| Vigilance state recognition is an important research direction in the field of braincomputer interface(BCI)and an important subject in artificial intelligence research.The recognition of vigilance can help humans to understand their brain state and adjust their working state in real time.Especially for the workers who are engaged in certain risks,the recognition of vigilance can effectively improve work efficiency and reduce the accident rate and protect people's lives and property safety.The main research contents of this article include:(1)A set of experimental scheme was designed based on the existing real multimodal trial signal acquisition system.Volunteers were recruited to complete specific tasks,and the Electroencephalogram(EEG)and functional near infrared spectroscopy(fNIRS)signals generated during the task were recorded.(2)EEG and fNIRS signal preprocessing: We used the independent analysis algorithm(ICA)to denoise and separate EEG signals from the frontal electrode signals.The fNIRS signal is filtered by a 0.5Hz low-pass filter and a 0.05 Hz high-pass filter,and then calculated by the Beer-Lambert law to obtain concentration change signals of oxyhemoglobin(Hb O)and deoxygenated hemoglobin(Hb R).(3)Based on the EEG data set published by Shanghai Jiao Tong University,we construct a light convolutional neural network(LCNN).(4)Transfer the LCNN : in order to improve the accuracy and generalization of detection method of vigilance state,We construct a set of the LCNN transfer model with the dataset of Shanghai jiaotong university.Then we transfer our model to the selfcollected dataset.The transferred model achieved 95% accuracy.The experimental results show that our algorithm can train a good model based on a small amount of EEG+fNIRS signals,compared with 90% accuracy of the traditional machine learning,we can obtain an accuracy of 95%.To a certain extent,it can solve the current problem of BCI Equipment which the equipment is too complex. |
PDF Full Text Request