We introduce 3-layer adaptive system, which uses known techniques in human detection in layer 1 (such as NN and HOG), and pass the results (human area) into layer 2, which extracts the contour and deforms it into its parts in order to estimate the human pose, using a knowledge base and mathematical equations that constraint the natural human poses, and assign correctness parameters to every result depending on constraints satisfactions. after that, the results and parameters are sent to layer 3 that holds the statistical information about the system work, and constructs the 3D model of the body, and controls each method weight due to its effectiveness.