Motion Sensors Based Human Arm Pose Estimation

Abstract

The elements of a human arm, forearm to upper arm, execute a series of motions to perform certain actions during daily activities. The performed actions are the result of combining the flexion-extension, abduction-adduction in different orders and with different angles. To capture the motions and use them for different applications, the angles for the arm's elements are needed for all three planes, coronal, transversal and sagital, during the motion. The proposed prototype has three MPU6050 IMU sensors connected to an Arduino Uno development board, through a TCA9548A demultiplexing chip, which transmits the captured motion data to a computer that uses MATLAB to post-process the data and simulate the motion in a 3D environment. The system has a reference situated on the spine and two points along the human arm, to capture its motion, the first in the middle of the Humerus bone, and the second is in the middle of the forearm. We've proposed this prototype, not only for accuracy, cost-effective hardware-wise, but also, operating it, represents and advantage, allowing flexibility and ease of use. This prototype has the capability not only to capture the human arm's motion, in real time, but also to discover the natural limits of the motions which a human arm can do. The sets of motion, that follow daily routines for motion, of a person, can be integrated in an exoskeleton's control system, to reproduce scenarios with the help of which possible patients can train themselves, or use in therapy, in order to regain mobility of their arms, that were formerly affected by neuromuscular diseases or physical injuries of the muscles or bones.