IMU-based motion capture system for real-time body joint angle measurement

A precise estimation of joint angles of the human body is an important requirement for assistive robotic systems. Real-time motion capture allows human-robot interactions with the upper limb during sophisticated positioning tasks. The measured joint angle trajectories provide a control variable for robots that are used in the area of rehabilitation, assistance or telemanipulation. Common camera based motion capture systems for movement analysis are expensive and require a stationary installation. Several commercial products use inertia measurement units (IMUs) for mobile motion capture of the human joint angles. However, costs of several thousand euros prevent the application in a broader field of medical robotics. In this work we present a low-cost motion capture systems for joint angle estimation of the human body. The modular system is based on MEMS IMUs (type BNO055, Robert Bosch GmbH, Germany) with an intelligent 9-axis absolute orientation sensor, which includes system on chip sensor fusion and filtering. Therefore, an additional Kalman or complementary filter is not required. For motion capture of the upper limb of a human user, four IMUs are fixed to the torso, upper arm, lower arm and hand with elastic bands. An additional microcontroller reads quaternions of the IMU sensors with 100 Hz sample rate via I2C to compute the rotation matrix of each sensor element. After all IMUs are calibrated, the relative position of each element is computed with an analytic procedure, using gravity-vector and omega calibration. The overall calibration process is done in less than ten seconds by moving the arm back and forth for a few times. The RMSE of the measured joint angles is smaller than two degrees for static positioning and five degrees for dynamic movements during activities of daily living. We successfully use the developed system to control an upper limb exoskeleton for robot-assisted rehabilitation