Motion Retrieval Based on Multiple Instance Learning by Isomap and RBF
Jian Xin Xiang · 2007
In this paper, a new learning method is proposed for human motion data analysis. In order to train motion data by the method of multiple instance learning, each human joint's motion clip is regarded as a bag, while each of its segments is regarded as an instance. Due to high dimensionality of motion's features, Isomap nonlinear dimensionality reduction is used. An algorithmic framework is used to approximate the optimal mapping function by a Radial Basis Function(RBF) neural network for handling new data. Then data driven decision trees based on multiple instance are automatically constructed to reflect the influence of each point during the comparison of motion similarity. Some experimental examples are given to demonstrate the effectiveness and efficiency of the proposed algorithm.