Estimation of 3-D human body posture via co-registration of 3-D human model and sequential stereo information

Abstract

In this paper, we present a technique for estimating three-dimensional (3-D) human body posture from a set of sequential stereo images. We estimated the pixel displacements of stereo image pairs to reconstruct 3-D information. We modeled the human body with a set of ellipsoids connected by kinematic chains and parameterized with rotational angles at each body joint. To estimate human posture from the 3-D data, we developed a new algorithm based on expectation maximization (EM) with two-step iterations, assigning the 3-D data to different body parts and refining the kinematic parameters to fit the 3-D model to the data. The algorithm is iterated until it converges on the correct posture. Experimental results with synthetic and real data demonstrate that our method is capable of reconstructing 3-D human posture from stereo images. Our method is robust and generic; any useful information for locating the body parts can be integrated into our framework to improve the outcomes.

Principal direction analysis-based real-time 3D human pose reconstruction from a single depth image

“…The joints of the model are connected with a kinematic chain and parameterized with rotational angles at each joint [21,20]. Our 3D synthetic human body model is defined in the 4-D projective space as…”

“…In [21], the authors developed a new algorithm based on expectation maximization (EM) with two-step iterations: namely, body part labeling (E-step) and model fitting (M-step). The depth silhouette and the estimated 3D human body model of this method were represented by a cloud of point in 3D and a set of ellipsoids, respectively.…”

“…In this paper, to overcome the limitations of the previous approaches [21,19], we propose an improved real-time 3D human pose estimation algorithm based on Principal Direction Analysis (PDA) of each recognized body part from a single depth image. In our work, human body parts of the depth silhouette are first recognized via the trained RFs with our synthetic training DB and RFs [3].…”

An embedded human motion capture system for an assistive walking robot

“…The objective of the segmentation step is the extraction of the 3D data related to each human limb. Unlike other approaches in the literature [18], our 3D human body model fitting is decomposed in two steps in order to reduce the cpu-time consuming and to avoid numerical instabilities 1 . Then, the Gauss−Newton algorithm is used to fit each limb by a cylinder.…”

Accurate 3D Human Pose Recognition via Fusion of Depth and Motion Sensors

“…In the presented methodology, we reconstruct human body poses using a single depth camera for a whole body as done in [5], [6]. In addition, we acquire information of the body limb (i.e., position and twist information of arms or legs) from two IMUs which are separately attached onto the right and left wrists or legs.…”