Low-cost Motion Capture Suit using Inertial Sensors

Motion capture technology has now existed for several decades and has been used in many different fields for a variety of purposes. In the entertainment industry, it has facilitated the creation of realistic and complex animations that, if created manually, would be too difficult and time-consuming. Different motion capture technologies have been invented over the years and they all have their strengths and weaknesses. Inertial motion capture is a low-cost alternative that relies on inertial sensors to estimate the orientation and position of a tracked object in three-dimensional space. Inertial sensors are a combination of a three-axis gyroscope and a three-axis accelerometer and are often contained in devices called inertial measurement units. In recent years they have become smaller, more lightweight, cheaper, less power-consuming and offer high sampling rates which makes them ideal for building a motion capture system. However, the outputted measurements from these sensors suffer from distortion which means that there needs to be a calibration procedure in place in order to minimize these distortions from the measurements. In this work, I developed a completely wireless configurable inertial motion capture solution with a robot-assisted calibration procedure. This motion capture solution con-sists of multiple motion trackers that are attached to the capture subject and wirelessly transmit the motion data to a receiving computer. I implemented a quaternion-based Extended Kalman filter as a sensor fusion method that uses the inertial data to estimate the orientation of the motion tracker. Due to the absence of a magnetometer sensor in this tracking solution, it is difficult to maintain a good enough accuracy when estimating the yaw angles of the motion trackers which leads to accumulated drifting errors over time. Therefore, this solution is only suitable for recording short animations for humanoid 3D characters. Furthermore, I compared my developed motion trackers to an existing commercially available tracking solution and the results indicate, with the exception of the low accuracy of the yaw angle estimation, acceptable orientation estimates.