Pnadilha et al. [7] proposed a method for sensor calibration. In the study, a 3-axis accelerometer, a 2-axis gyroscope and Kinect are used for joint angle estimation.
Online calibration was developed based on Kalman filtering to combine the data from both inertial sensor and Kinect to improve the overall estimated motion. Many research studies reported that sensor-based tools for rehabilitation is still not 100% accurate, however, an error of ±5° in goniometry measures is suggested as an acceptable range for several clinical circumstances [8].