Risk Analysis for Fall Detection: Exploiting Using GAIT, Part Affinity Field and Machine Learning

According to our survey, the primary cause of injury or death for people around 60 years are from falls (Christiansen TL et al., in Jt Comm J Qual Patient Saf 46(22), 129–135 (2020); Grossman DC et al., JAMA 1696–1704 (2018)). It is estimated that around 35% of elder get injured by falling every year, scientifically, it is proven that the reason for falling is because of an imbalance of the center of gravity, i.e., the CG (center of gravity) of the person is unstable. It is necessary to find fast and effective way to find fall detection to help the elderly fall. This fall detection can also be used for patients in hospitals and in road traffic for drunk people. In this paper, we are going to detect the fall of a person using an open pose model. Basically, we are going to extract the key point/joints of a person and then we are going to analyze the gait of a person by taking the centroid as a key point for fall detection. This method is effectively able to detect falls with an accuracy of 90%.