Multi-Scale Virtual Human Modeling Method

With the abroad application of ergonomics and the fast development of interdisciplinary fields, virtual human modeling is realized to be an effective tool for studying human relevant issues. More and more emphasis has been placed onto human modeling through the use of anthropometric data of the target population to reveal their features of body shape and size. Modern anthropometric human modeling is based on multivariate methods. However, these concepts of human modeling and its relevant software have a same limitation: pre-determined body dimensions, in terms of number and items, are used. It specifies the same static model input, without considering different model requirments and to what extent a human model is actually needed. This study developed a new idea to dynamically provide the users with multi-scale human models according to the requirments.This method was to provide different key body dimension sets in terms of number and items based on the correlation between body dimensions, and use them as the model input to build up the entire body model. Hierarchical factor analysis, developed from recursively using factor analysis, was employed to discover the detailed correlation, classify body dimensions and present them in tree-like plot; all the dimensions in the same group were ranked according to the combined consideration of factor loading and variable variance; the required number of key body dimension was allocated onto groups by the variance explained; users could choose key body dimensions based on their need and finally obtain the entire body model through stepwise linear regression. A human body modeling system was built up by MATLAB programming. Its modeling function could satisfy users with a required final human model through this multi-scale method, while its analyse function focused on the validation of the model's and the specific dimension's goodness of fit and discussion on combined weight w.ANSUR88 database, HQL database and 24 databases from HSIAC were used as case studies with the discussion of their results.