A method for customizing 3D virtual human body models based on Multi-class Support Vector Machine

Abstract

How to generate a personalized 3D virtual body model conveniently and quickly is playing an increasingly important role in computer animation, virtual reality, entertainment, e-commerce and many other areas. Some related researchers just simply adjust human characteristic parameters to generate body model using existing 3D body model. In this article, in order to generate the personalized 3D virtual body model quickly, an approach based on Fuzzy Support Vector Machines (FSVM) is suggested. This constructs a classification model of the personalized body characteristic parameter. The one-versus-one (OVO) method based on the binary tree is used to handle a multiclass problem by breaking it into various two-class problems. Application of the method shows that the method of FSVM has the characteristics of less calculation and less error in the allowed range than the classical neural network.