Driving Safety Monitoring System for Personal Mobility by Deep Learning

This study proposes a system for monitoring personal mobility driver's careless driving caused by visual and auditory cognitive load occur during driving personal mobility. The N-Back Task was executed on the driver, causing cognitive loads in the experimental section, and the vibration data generated at this time was transformed into images using 1)STFT or 2)Wavelet Transform(WT). Multimodal deep learning classified image from 6-axis sensor in normal driving section and cognitive load driving section with accuracy about 99%. The proposed system can monitor drivers to provide warning and automatic control of careless driving and improve the safety of nearby pedestrians.