Fatigue driving recognition based on deep learning and graph neural network. (July 2021)

• Record Type: Journal Article
• Authors: Lin, Zhiqiang; Qiu, Taorong; Liu, Ping; Zhang, Lingyun; Zhang, Siwei; Mu, Zhendong
• Is Part Of: Biomedical signal processing and control. Volume 68(2021)
• Publication Date: 2021-07
• Subjects: Electroencephalogram -- Fatigue driving -- Functional brain network -- Graph neural network -- Stability; Signal processing -- Periodicals; Biomedical engineering -- Periodicals; Signal Processing, Computer-Assisted -- Periodicals; Image Processing, Computer-Assisted -- Periodicals; Biomedical Engineering -- Periodicals; 610.28
• DOI: 10.1016/j.bspc.2021.102598
• Languages: English
• ISSNs: 1746-8094

Abstract

Highlights: The proposed CNN denoise network estimates the noise distribution in EEG signal, which can process EEG signals with different signal to noise ratios. The proposed functional brain network construction method dynamically constructs functional brain network to represent brain activity. The proposed graph neural network model recognizes fatigue driving by combining functional brain network with EEG features. Abstract: The performance of traditional fatigue driving recognition model is greatly reduced under noisy environment. In order to solve this problem, a fatigue driving recognition framework which could denoise the EEG signals based on deep convolutional neural network and dynamically constructed functional brain network was proposed. The framework constructed functional brain network based on singular value entropy and fractal dimension and used a graph neural network based on B-spline curve spatial convolution to classify the functional brain network. Recognition accuracy and recognition stability of the framework were analyzed. Experiment results show that our framework can achieve high accuracy in fatigue driving recognition. Furthermore, our framework can maintain the high accuracy of the recognition effect when the EEG channels are affected by noises.