Fall Detection with a Single Commodity RGB Camera Based-on 2D Pose Estimation

Abstract

This study is a joint gesture estimation algorithm with LSTM to train the device to recognize a drop with a single non-modified RGB camera. Additionally, we try to detect as soon as the person falls. (Not when he / she is already lying on the ground) This work focuses on optimizing the modeling while maintaining computational resources for the real-time application.In older adults, falls can cause severe injuries. Even right before dropping, reaction  time is important, so the near-by device can respond on time. This article introduces a method  for early detection of falls that is non-invasive. Our method only uses a single RGB Commodity  Camera. With CNN-based, we estimate human-position, then use the pose to recognize falls  without any calibration and additional equipment, processing join data and predict with LSTM  model. The model achieves 97% in accuracy. It can be extended to a smart home. Information obtained testing the system, and many supports for the  course of the study.  Will benefit fall Detection with a Single RGB Camera