Fast tool to evaluate 3D movements of the foot-ankle complex using multi-view depth sensors

Realtime multi-person 2D pose estimation is a key component in enabling machines to have an understanding of people in images and videos. In this work, we present a realtime approach to detect the 2D pose of multiple people in an image. The proposed method uses a nonparametric representation, which we refer to as Part Affinity Fields (PAFs), to learn to associate body parts with individuals in the image. This bottom-up system achieves high accuracy and realtime performance, regardless of the number of people in the image. In previous work, PAFs and body part location estimation were refined simultaneously across training stages. We demonstrate that a PAF-only refinement rather than both PAF and body part location refinement results in a substantial increase in both runtime performance and accuracy. We also present the first combined body and foot keypoint detector, based on an internal annotated foot dataset that we have publicly released. We show that the combined detector not only reduces the inference time compared to running them sequentially, but also maintains the accuracy of each component individually. This work has culminated in the release of OpenPose, the first open-source realtime system for multi-person 2D pose detection, including body, foot, hand, and facial keypoints.

A simple external marker system and algorithms for computing lower extremity joint angle motion during level walking were developed and implemented on a computer-aided video motion analysis system (VICON). The concept of embedded axes and Euler rotation angles was used to define the three-dimensional joint angle motion based on a set of body surface markers. Gait analysis was performed on 40 normal young adults three times on three different test days at least 1 week apart using the marker system. Angular motion of the hip, knee, and ankle joints and of the pelvis were obtained throughout a gait cycle utilizing the three-dimensional trajectories of markers. The effect of uncertainties in defining the embedded axis on joint angles was demonstrated using sensitivity analysis. The errors in the estimation of joint angle motion were quantified with respect to the degree of error in the construction of embedded axes. The limitations of the model and the marker system in evaluating pathologic gait are discussed. The relatively small number of body surface markers used in the system render it easy to implement for use in routine clinical gait evaluations. Additionally, data presented in this paper should be a useful reference for describing and comparing pathologic gait patterns.

Biomechanical analysis is a powerful tool in the evaluation of movement dysfunction in orthopaedic and neurologic populations. Three-dimensional (3D) motion capture systems are widely used, accurate systems, but are costly and not available in many clinical settings. The Microsoft Kinect™ has the potential to be used as an alternative low-cost motion analysis tool. The purpose of this study was to assess concurrent validity of the Kinect™ with Brekel Kinect software in comparison to Vicon Nexus during sagittal plane gait kinematics. Twenty healthy adults (nine male, 11 female) were tracked while walking and jogging at three velocities on a treadmill. Concurrent hip and knee peak flexion and extension and stride timing measurements were compared between Vicon and Kinect™. Although Kinect measurements were representative of normal gait, the Kinect™ generally under-estimated joint flexion and over-estimated extension. Kinect™ and Vicon hip angular displacement correlation was very low and error was large. Kinect™ knee measurements were somewhat better than hip, but were not consistent enough for clinical assessment. Correlation between Kinect™ and Vicon stride timing was high and error was fairly small. Variability in Kinect™ measurements was smallest at the slowest velocity. The Kinect™ has basic motion capture capabilities and with some minor adjustments will be an acceptable tool to measure stride timing, but sophisticated advances in software and hardware are necessary to improve Kinect™ sensitivity before it can be implemented for clinical use.