Motion Control and Retrieval Based on User Feedback Information

摘要

The spatial and temporal features of user feedback information in 3D Motion control and retrieval are analyzed, based on a novel user feedback learning method. The method treats the positive and negative examples returned at different rounds with different weights in the learning process according to the spatial and temporal featuress of the relevance feedback information. With these features, the proposed SKSVM (Temporal and Spatial-featured Support Vector Machine) method improves the performance of the learning significantly as compared with the conventional SVM method.