Fall Detection System for Depression Angle RGB Image based on Human Skeletonch

In recent years, Taiwanese society is gradually aging, and the existence of fall detection systems is crucial to protect the health and safety of the elderly in their daily lives. In this paper, we used the Openpose human skeleton recognition algorithm to extract the 2D human skeleton information from RGB images and took actions to recognize the 2D human skeleton information by the LSTM. In order to solve the problem of incomplete and inaccurate human skeleton information extracted from the top angle camera, this paper proposes an algorithm that uses an object recognition model to assist in extracting human skeleton information to improve its skeleton recognition rate, and then enhance the accuracy of the subsequent LSTM model to obtain more accurate fall detection results.