A Review on Deep Learning-Based Fall Detection Systems: Impact of Visual Quality and Motion Dynamics on Time Accuracy

Elderly falls are one of the most significant issues in the area of public health because of the serious consequences on health and independence. Deep learning-based fall detection systems have become the hotspot due to their capabilities of detecting falls with increased accuracy over the traditional way. Nevertheless, they are frequently affected by the environmental and visual factors. The role of the independent variables, namely brightness, motion speed, and image noise, which can influence system reliability, and time accuracy as the dependent variable, are discussed in this review. Considering the recent research, the issue of adaptive image enhancement can enhance the fall detection in low-illumination conditions, which proves the significance of brightness in the recognition. On the same note, motion dynamics and image quality are also vital in the development of real time performance. This article summarizes recent results, pinpoints the methodological issues and outlines future research into strong and adaptive deep learning-based fall detection systems that can assist the elderly in a variety of settings.