A Three-Dimensional Animation Character Dance Movement Model Based on the Edge Distance Random Matrix

Abstract

In this paper, we use the edge distance random matrix method to analyze and study the dance movements of 3D animation characters and design a 3D animation character dance movement model. Firstly, each dance movement video in the dataset is divided into equal parts, while the segmented videos are subjected to the operation of accumulating edge features separately, and the edge features in all video images within each segment are accumulated into one image, and the directional gradient histogram features are extracted from them. Finally, a set of directional gradient histogram features are used to represent the local appearance and shape of the video dance movement. The reconstructed human movements are obtained mainly by fitting the 3D human coordinates in the image space using the human model and the estimated depth coordinates, which are currently more mature and are chosen from existing techniques. In other combinations, the performance of the method in this paper is better than the recognition results of the benchmark method, especially when the similarities of dance movements in the towel flower combination and the piece flower combination are too high. In response to the problem of the reconstructed action appearing to have ground penetration, slipping, and floating feet, a method is proposed to optimize the fitting human model foot problem according to the foot touching the ground. The experimental results show that the method can replace the traditional motion capture method to a certain extent, simplify the use of motion capture, and reduce the cost of motion capture. In the model deformation stage, to reduce the deformation quality problem during the model motion and improve the efficiency of weight calculation, a model deformation method combining double quaternions and bounded double tuning and weights are given. The static 3D model is tetrahedralization, and then the skeletal control points of the model are set and the weight of each skeletal segment to the model is calculated; next, the 3D model is mixed and bound with the skeletal data using the dual quaternion skinning algorithm; finally, the static 3D model motion is driven by the skeletal data. Experiments demonstrate that the method results in better deformation of the 3D model during rotation.