Human pose estimation is defined as the joints localization of a person or a group of people in an image or a video. Nowadays, pose estimation es commonly used by detection systems to monitor patients in hospital environments. However, is not an easy task because of the need of specialized personnel to manually evaluate the human posture, or using special equipment like: e-health devices(watches, strips, handles), markers or high cost cameras to control a limited space. In that order, this study proposed a performance evaluation of computer vision models used with videos captured with 3 different devices to develop a low cost pose and fall estimation system which is adaptable to different scenarios. As a result it was obtained a precise system that does not need the use of sensors or a high cost set of cameras, and which methods decrease the pose and fall detection time.
