Application of a BiLSTM Model for Detecting Driver Distraction Caused by Hand-Held Mobile Phones, Utilizing Physiological Signals and Head Motion Data

Driver distraction from the use of mobile phones significantly contributes to road accidents This research introduces a Bidirectional Long Short-Term Memory Network (BiLSTM) deep learning model that utilizes drivers' physiological responses and head motion data to detect driving distraction caused by handheld mobile phone usage. This study utilized a simulation-based approach, employing a motion platform and simulated environment to replicate real driving conditions. Physiological signals, including heart rate, breathing rate, galvanic skin response and skin temperature were collected using a body worn sensor, while head motion data were captured using accelerometer and gyroscope sensors. This study is novel as it employs motion platforms to explore how phone distractions affect physiological reactions and head movement data, specifically when integrating advanced deep learning methods for detection purposes. Through rigorous training and optimization of the BiLSTM model's hyperparameters, the research achieved an accuracy rate of 99.78%. Additionally, precision, recall, and F1-score were reported at 99.74%, 99.82%, and 99.78%, respectively. The findings illustrate the effectiveness of the developed BiLSTM model in accurately predicting driver distraction through the analysis of physiological signals and head movements. The success of this model suggests its potential application in future driver-assistance systems to alert and notify drivers, thereby reducing the likelihood of road crashes caused by mobile phone-related distractions.