A new fused feature descriptor based on heat kernel and wave kernel signatures is proposed to calculate shape correspondences between isometric 3D models. First

the Laplace operators of 3D models are calculated to obtain eigenvectors and eigenvalues

which are defined as the basic parameters for computing the heat kernel and wave kernel signatures of the source and target models

respectively. Second

heat kernel and wave kernel signatures are fused together as a new feature descriptor

which serves as a constraint for points sampled uniformly in random. Finally

the shape correspondences between models are calculated by a minimum-value matching algorithm. Experimental results show that the accuracy of the correspondence ratio calculated using the proposed feature-descriptor constraint increases by 19.429% and 4.857% on average with respect to the correspondence ratios calculated using heat kernel signature and wave kernel signature constraints

respectively. The fused feature descriptor presented in this article is applicable to calculating correspondences between isometric 3D models or approximately isometric 3D models. Compared with the descriptors that only use heat kernel or wave kernel signatures

the fused-feature descriptor was able to obtain more accurate correspondences.