Wearable sensor systems have the potential to offer advancements in the study of motion disorders, particularly outside of a laboratory setting during activities of daily living or on a football field. Advantages like portability and the capability to gather real-world data have resulted in the rapid adoption of these sensors in various studies for gait analysis, balance control evaluation, physical activity recognition and fall prevention. However, before using wearable sensors in long-term acquisition studies, it is necessary to quantify and analyze errors and determine their sources. In this study, the accuracy of joint angles and velocities measured with the wearable inertial measurement unit (IMU) sensors were compared to both measurements from an optical motion-tracking system and from encoders on a robotic arm while it completed various predetermined paths. The robotic arm uses incremental encoders at each joint to measure and calculate its Cartesian motion relative to a reference frame using inverse kinematics. Motion profiles of the robotic arm were tracked using the onboard encoders, an eight-camera Vicon (Oxford, UK) motion-tracking system with passive retro-reflective markers, and four wearable IMUs by APDM (Portland, OR). In order to better isolate various types of contributing errors, linear, planar, and 3-dimensional robot motions were used. Data were collected from the sensors over several hours, which provided insight into time-based effects as well as management of large amounts of data for future long-term tracking applications. In addition, the authors have previously seen acquisition errors with high-speed gaits, thus robotic arm trajectories of varying velocities were used to provide further insight into these rate-based effects. Angular velocity and joint angles were compared for all three systems and used to investigate the hysteresis, drift and time-based effects on the IMUs as well as their accuracy during motion tracking. Effects on IMU performance due to the application of filtering algorithms were not investigated. The results show that the IMUs were able to calculate the joint angles within a clinically acceptable range of the gold standard optical motion-tracking system. The IMUs also provided accurate trajectory recognition and angular velocity measurements relative to the known motion input of the robotic arm. Future work will include the development of algorithms to detect gait abnormalities such as those seen in patients with mild traumatic brain injury (mTBI). To complement human subject testing with gait pathology, controlled introduction of gait deviations into this robotic testing framework will allow for well-characterized unit testing, providing more robust algorithm development.
A Low-Cost Motion Capture System for Small-Scale Wave Energy Device Tank Testing
