Digital video production is increasing day after day and the effectiveness and efficiency of the image recognition mechanism has became increasingly important. In this study, a new skeleton-based shape recognition algorithm, which is generated by drawing spheres with increasing radii around the skeleton, is presented. After this operation, drawn spheres partially remain the inside of shapes because each frame corresponds to the center of a maximum tangent sphere. The ratio between number of pixels of the drawn sphere that is remaining in the shape, and the total number of pixels of the drawn sphere is used to identify skeleton similarities and shapes. The experimental evaluation of a comparison with previous techniques and an approach that suggested to show the effectiveness and reliability of the operation is presented.