Deep learning method for risk identification under multiple physiological signals and PAD model
Document Type
Article
Publication Date
2-1-2022
Abstract
The number of vehicle ownership keeps increasing with the fast development of China's society and economy. Consequently, there are surging road traffic accidents, which seriously endanger the safety of people's lives and property. Hence, the present work aims to reduce the probability of traffic accidents and improve the efficiency of driver risk identification. The driver is taken as the research sample. Driver's physiological and Pleasure Arousal Dominance (PAD) data are collected using virtual driving equipment. Recurrent Neural Network (RNN) is applied to identify multiple physiological signals. Support Vector Machine (SVM) is utilized to classify the data. The performance of the proposed model is validated through data analysis of multiple volunteers. Results demonstrate that the PAD physiological signals can improve the rate of drivers recognizing risks. In the meantime, these signals help to understand the driver's emotions, thereby identifying the risks faced by the driver. By combining SVM and RNN, the accuracy of risk identification reaches 91.44%, which to some extent proves the effectiveness of the proposed model. The results have vital reference value for reducing traffic accidents and ensuring driving safety.
Keywords
Multiple physiological signals, PAD model, Risk identification, Support vector machine, Recurrent neural network
Divisions
fac_eng
Funders
None
Publication Title
Microprocessors and Microsystems
Volume
88
Publisher
Elsevier
Publisher Location
RADARWEG 29, 1043 NX AMSTERDAM, NETHERLANDS