A Survey on Vision-Based Elders Fall Detection Using Deep Learning Models

Falls are a leading source of severe wounds in adults over the age of 60, according to WHO. It has the potential to cause significant wounds, and impairments, and can even lead to demise, particularly in older citizens who live alone. As a result, we can improve these people’s quality of life by implementing different automated fall detection systems. Using multiple deep learning architectural models, this study provides a full-scale evaluation of recent fall detection approaches. Its goal is to serve as a resource for academics and industry to research various fall detection methods. We investigated nearly all contemporary along with promising deep learning approaches for fall tracking and detection and divided them into two classes: convolutional neural network-based (CNN) systems and recurrent neural network-based (RNN) systems. The focus of this research is to offer a contrast of several deep learning-based fall detection systems like CNN, LSTM, and RNN as well as illustrated a comprehensive table that gives an overview of the paper published from the year 2017 to 2021 that could be used in better understanding the role and significance of the systems in the fall detection task. At the closure of the report, obstacles and further development have been mentioned. This survey can assist scholars in better understanding of present systems and proposing new techniques by addressing the challenges that have been identified.