This paper addresses the problem of recovering 3D human pose from single 2D images using Sparse Representation. While recent Sparse Representation (SR) based 3D human pose estimation methods have attained promising results estimating human poses from single images, their performance depends on the availability of large labeled datasets. However, in many real world applications, accessing to sufficient labeled data may be expensive and/or time consuming, but it is relatively easy to acquire a large amount of unlabeled data. Moreover, all SR based 3D pose estimation methods only consider the information of the input feature space and they cannot utilize the information of the pose space. In this paper, we propose a new framework based on sparse representation for 3D human pose estimation which uses both the labeled and unlabeled data. Furthermore, the proposed method can exploit the information of the pose space to improve the pose estimation accuracy. Experimental results show that the performance of the proposed method is significantly better than the state of the art 3D human pose estimation methods.