Abstract
With the increasing significance of smart transportation in China, there is a growing interest in the urbanization process to ensure its safe operation. This study introduces a driver behavior detection system combining an electromagnetic-triboelectric wristband as a self-powered sensor with deep learning techniques. The system consists of the electromagnetic power generation module, the triboelectric nanosensor module, and the driver behavior detection module. The sensor includes an eccentric pendulum with magnets and polytetrafluoroethylene beads, a shell with coils at the bottom, and copper rings on both sides of the inner wall. During the rotation of the eccentric pendulum, the magnetic field lines cut the coils to generate electricity, while the polytetrafluoroethylene beads roll and rub against the copper ring to produce electrical signals. Experiments were conducted to evaluate the effectiveness of power generation and sensing, demonstrating a power density of up to 26.78 W/m3. The signals for six driving behaviors were collected, followed by preprocessing and dataset production. The proposed multi-scale convolutional recurrent neural network model was trained, tested, and compared against traditional time-series signal classification models, and its effectiveness was verified. The accuracy of the test set reached 98.21%, demonstrating the feasibility of the proposed driver behavior detection system and its application potential.
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have