An Explainable AI Framework for Vision-Based Human Fall Detection in Fog Infrastructure

The activity recognition using visual data is the primary emphasis of this study. One of the most significant issues that poses a threat to the lives of older people is human fall. It is possible for elderly people to suffer irreparable disability or even pass away as a result of accidental falls. Within the realm of healthcare, fall detection has emerged as a significant research topic, necessitating the development of systems that are both more dependable and efficient to intelligently categorize fall actions. Continuous monitoring of old individuals has become practical because of the growth of the Internet of Things, which includes the creation of wearable sensors, ambient sensors, and cameras. Using Deep Learning classification and a Convolutional Neural Network with three hidden layers, the work that is being offered assures the detection of falls. In order to assess the effectiveness of the Deep Learning model, the Fall Detection Dataset is used. Additionally, for intelligent classification systems to be accepted by healthcare practitioners, they need to be trustworthy. Explainable artificial intelligence models, such as LIME and SHAP, are tested in this study in order to explain the categorization of fall activity. Because of defining the limits of the input picture, the outputs of XAI models demonstrate the feature that is responsible for prediction.