Kalman Filter-Based Noise Reduction Framework for Posture Estimation Using Depth Sensor

Significant benefits can be achieved through the integration of cognitive computing technologies in healthcare delivery services, including physiotherapy. The traditional approach to physiotherapy requires attaching a marker-based tracking device with the patients and conducting analysis to diagnose by physiotherapists and chiropractors. Tracking efficiency of patient rehabilitation frequently depends on the physiotherapist's notes, which is tedious and prone to errors. In order to streamline the process of data collection and record-keeping, and to make more informed decisions on the effectiveness of prescribed therapy, depth sensors can be integrated with current physician practices. This paper is one of the very first attempts to assist physicians through proprietary Kinect sensor-based technologies. The goal is to make sure static posture estimation is highly accurate. Thus, this paper introduces the solution through a noise reduction framework where the Kalman filter and a recursive noise reduction algorithm are combined to improve the accuracy and the consistency of the human 3D skeleton motion data. The Kalman filter is used for the reduction of tremors by abnormal estimation of body joints in real-time using Kinect v2. The posture correction algorithm is incorporated in the proposed framework to reduce anthropometrically inconsistent estimation of limb lengths of the human body. The proposed posture correction algorithm was extensively validated on the proprietary data set.