Optimization-based posture reconstruction for digital human models

Highlights

• Optimization-based formulation for posture reconstruction is developed.

• Local optimization solution methods are computationally efficient.

• The local optimization solution method 3 is the best choice for posture reconstruction.

• Z-test was carried out to determine the difference of each solution method.

• Probabilities are calculated for each joint center within a given tolerance.

Abstract

Digital human modeling provides a valuable tool for designers when implemented early in the design process. Motion capture experiments offer a means of validation of the digital human simulation models. However, there is a gap between the motion capture experiments and the simulation models, as the motion capture results are marker positions in Cartesian space and the simulation model is based on joint space. Therefore, it is necessary to map the motion capture data to simulation models by employing a posture reconstruction algorithm. Posture reconstruction is an inherently redundant problem where the collective distance error between experimental joint centers and simulation joint centers is minimized. This paper presents an optimization-based method for determining an accurate and efficient solution to the posture reconstruction problem. The procedure is used to recreate 120 experimental postures. For each posture, the algorithm minimizes the distance between the simulation model joint centers and the corresponding experimental subject joint centers which is called the mean measurement error.

Introduction

Digital human modeling plays an important role in human-centric design as it allows the designer to study the interaction between the user and the product. A digital human, in a general sense, is a software application that aims to replicate the human in form and function. A typical advantage of incorporating digital humans into the design process is that it reduces the number of iterations for design or the need for experimental testing, thus reducing overall costs and time to market. A typical digital human simulation environment allows a designer to test a product in a virtual environment with the digital human providing the necessary feedback, such as reachability, on the product or activity. This allows the designer to refine the design before a physical prototype is built. Since the goal of a digital human simulation is to replicate humans as closely as possible, it is important that all simulations are properly validated. Motion capture experiments can provide the tool for validating digital human models.

A typical optical motion capture experiment involves placing reflective markers on a subject’s skin and clothing and capturing the marker movement with high-speed cameras. As the subject performs a physical task, the marker position is tracked, and a history of each marker’s position throughout the experimental trial is recorded. The position data for each marker is tracked in Cartesian space. To validate a digital human simulation, the experimental data and simulation data are compared for a given physical task. If the digital human simulation does not replicate the experiment, the simulation can be modified. Once a simulation has been validated, the simulation can replace the experiment altogether, allowing future experimentation and design to be completed wholly in the virtual realm. However, there is a gap between the motion capture data and simulation output.

Data from motion capture are in Cartesian space and the digital human simulation models are in joint space. Therefore, one has to map the data from an experiment to the simulation model by using posture reconstruction. There are several techniques for posture and motion reconstruction available in the literature. Posture and motion reconstruction problems can be classified as over-guided, exactly-guided, and under-guided (Ausejo & Wang, 2009). Under-guided problems arise when there is not sufficient motion data to determine the position and orientation of each portion or segment of the human model. Exactly-guided problems occur when there are an equal number of human model degrees of freedom and measured marker coordinates. Exactly-guided problems are rare and can be solved the same as over-guided problems. Over-guided problems, such as the posture reconstruction problems in this study, appear when there are more markers than strictly needed to capture the motion of the human or there is redundant information to define the posture of the human at each time frame. Over-guided problems can be further categorized into local-IK (IK-inverse kinematics) methods and global-IK methods (Ausejo & Wang, 2009). Local-IK methods are applied if joint constraints are not required or if there are only a small number of body segments studied. Global-IK methods are generally optimization-based solutions where the global measurement error, defined as the sum of the distance square between the measured and model marker positions, is minimized (Ausejo & Wang, 2009). Examples of global-IK methods can be found in the literature (Ausejo, 2006, Ausejo et al., 2006, Baerlocher and Boulic, 2004, Bodenheimer et al., 1997, Cerveri et al., 2005, García de Jalón and Bayo, 1994, Lu and O’Connor, 1999, Monnier et al., 2007, Riley et al., 2000, Roux et al., 2002, Wang et al., 2005, Zhao and Badler, 1994). For further information on other global-IK methods please see previous work (Gragg & Yang, 2011). When we develop physics-based posture prediction methods such as those presented in Howard, Cloutier, and Yang (2012), it is necessary to provide an accurate and efficient solution to the posture reconstruction problem for validation purposes. This paper proposes an optimization-based procedure for solving a posture reconstruction problem classified as an over-guided, global-IK method with relative coordinates.

A pilot study (Gragg & Yang, 2011) was carried out to determine a reliable posture reconstruction algorithm; however, it is desirable to develop a robust and generic optimization-based procedure. This generic method includes four local-optimum solution methods applied to 120 experimental postures. The optimal posture reconstruction pose is then determined by comparing the results from the four solution methods, with the optimal pose being the one where global measurement error is minimized. Error probabilities are also given for the solution method that produced the minimum global measurement error for all 120 postures.