Research Article
Articulated Human Motion Tracking Using Sequential Immune Genetic Algorithm
Algorithm 1
ISOMAP-based manifold reconstruction.
Input: The training data set . | Output: The mappings , and , . | Step 1: (preparing) | (1) Using the ISOMAP algorithm to compute the low-dim vector for the | original input vector ; | (2) Construct the following matrixes: | , , where is the | -neighbor of . | (3) Compute , where is the generalized inverse matrix of . | Step 2: (manifold reconstruction) | (1) Mapping from original space to latent space: , . | Given a high-dim pose vector , the corresponding low-dim vector can be computed as: | (1.1) Find the nearest neighbor of in , set it to be ; | (1.2) The low-dim vector correspondence to can be formulated as: | ; | (2) Mapping from latent space to original space: , . | Given a low-dim pose vector , the corresponding high-dim vector can be computed as: | (2.1) Find the nearest neighbor of in , represented as ; | (2.2) The high-dim vector correspondence to can be formulated as: | . |
|