A Robust Orientation Filter for Wearable Sensing Applications

Abstract

Advancements in the micro-electromechanical systems technology have enabled the realization of small-size, high-performance inertial motion and magnetic field sensors that are embedded in most modern-day smart gadgets. These sensors, when coupled with the high-speed computing and communication technologies may potentially enable in-home monitoring and assessment of human health in the forthcoming age of Smart home technologies, internet-of-thing, and internet-of-everything. However, because the sensor’s orientation is generally arbitrary, this may cause erroneous results of important health parameters such as gait speed and range of motion of the knee joint. Therefore, it is important that the sensor’s measurements be corrected for orientation. In this work, we designed, implemented, and validated a three-stage sensor fusion algorithm. A gradient descent approach was exploited to estimate the drift in and subtract it from the cumulatively integrated gyroscope data to obtain the orientation in real time. The roll and pitch angles were obtained from the first stage, whereas the second and third stages outputs a coarse and fine estimate of yaw angle, respectively. Since the estimation was obtained primarily from the gyroscope data, the estimated orientation was least affected by the external acceleration and magnetic disturbances. The performance of the proposed algorithm was validated with a publicly available dataset, and in presence of external acceleration and magnetic disturbances. Finally, some key gait parameters were derived from the gait measurements using the proposed filter that showed high conformity to the ground-truth values.