ViFa: an analytical framework for vision-based fall detection in a surveillance environment

The decrease in fertility rates and the increase in the average age of individuals are the main reasons behind the aging of the population. Challenges that come with an aging population include proper nursing care. Because the cost of healthcare is high and falls that cause injury or death in the elderly are escalating, they are a challenge for public welfare and research into reliable surveillance is essential. Non-intrusive fall detection systems are vital for reducing fall trauma and machine vision is an appropriate solution for detecting unusual events such as falls. Because there are varieties of vision-based fall detection (VBFD) methods and a comprehensive framework is lacking, comparison and evaluation of existing methods are difficult. In the current study, an analytical framework having three main components is proposed. First, existing VBFD methods are classified into three categories. Next, general evaluation criteria are defined for analysis of the proposed categorizations. Finally, each method is qualitatively analyzed using the proposed categorizations. The proposed framework can accurately provide identification and analytical comparison of existing methods. In addition, it can allow selection of the most appropriate methods and suggest improvements for existing methods.