An IoT-Based Hardware System for Road Accident Detection and Notification Using CNN-LSTM Algorithm

According to the World Health Organization (WHO), approximately 1.3 million people die each year globally, due to road accidents and the deaths have increased by 17% in 2021 compared to 2020. Fatalities and injuries caused by the accidents are now a matter of great concern. There are certain causes behind this but the most crucial issue is not receiving medical emergency services at time which eventually leads to a dead-end situation. In this scenario, a complete solution for automatic accident detection and notification is tile most needed. This does not only help to save millions of lives rather can create a positive impact on society. In this paper, we aim to develop an Internet of Things (IoT) based hardware system that can be installed in a car as an auxiliary device to collect data, detect an accident, and deliver notifications of the same with the exact location, to the relatives and the nearest emergency service providers. The device is composed of Raspberry Pi along with other sensors like an Accelerometer, Gyroscope, Global Positioning System (GPS), and also a Global System for Mobile Communications (GSM) module. To analyze and classify the collected data, the Convolution Neural Network and Long Short-Term Memory (CNN-LSTM) hybrid machine learning model is proposed. The model is trained with a time-series dataset and it predicts the occurrence with 92 % accuracy. The comparative performance of the proposed framework claims its supremacy over other models in terms of Accuracy, Precision, Recall, and F1-Scorp.