Covering of the Biomimetic Pattern Recognition of the High Dimensional Information Geometry

Objective: To research on the application of high dimensional space theory to the biomimetic pattern recognition. Procedures: The sample was constructed into a k-dimensional simplex in a high dimensional space and 0.85 times of the average distance between the vertices was chosen as the threshold value thus a convex cell body covering the k-dimensional simplex was constructed. By determining whether the sample point could be covered by the convex cell body derived from a certain sample group, the sample was recognized and classified. Methods: The orthogonal complementary space of the subspace was constructed and the Euclidean distance between the point and the subspace was calculated and the distance from the point to the k-dimensional simplex was further calculated. Results: The study solved the problem of whether the sample point is covered by a certain convex cell body and its correctness was testified via human-face recognition. Copyright © 2013 IFSA.