Fall Detection from Smart Wearable Sensors using Deep Convolutional Neural Network with Squeeze-and-Excitation Module

Older people are at a higher risk of injury or death due to falls than the general population. As the world's population ages, new fall detection and prevention technology must be developed quickly. The emerging technology sector is devoted to the development of tools that improve the lives of the elderly. The goal of a fall prevention system is to foresee and reduce the risk of a person falling. A fall detection system, on the other hand, tracks the fall. It alerts the user in the event of a fall to assist. The purpose of this study is to demonstrate the efficacy of fall detection utilizing wearable sensors. We developed the SE-DeepConvNet model, a hybrid deep convolutional neural network, to enhance the fall detection capability using squeeze-and-excitation modules. The proposed model uses wearable sensor data from the SisFall dataset, a publicly available benchmark dataset. We performed a series of experiments using various deep learning models, including the proposed approach. The experimental findings demonstrate that the introduced SE-DeepConvNet model performs better than other deep learning models (99.201% accuracy and 99.202% F1-score).