The Research On Theory Of Complicated Geometry Shapes In High-Dimensional Space And Its Application In Face Recognition

Because of the essential relation between information science andthe high-dimensional space, many issues and explanations in thehigh-dimensional space are corresponding to lots of process methods inthe information science, and the development of high-dimensionalgeometrical methods rises a new way on finding new directions of theinformation science. To solve the practical issues of pattern recognition,we study the theory of complicated geometrical shapes in high-dimensional space in this paper.Firstly, some concepts and axioms in high-dimensional geometry arepresented, and some distance algorithms that are used in patternrecognition are summarized. The method of Vertex Covering is alsoanalyzed. Then we analyze the distribution of face samples inhigh-dimensional space. The model of multi-degree of freedom neuronsis proposed based on the theories that are presented above. Lastly, the cognition algorithm based on multi-degree of freedom neurons is appliedto face recognition, and experiments prove its efficiency. The process ofthe experiment show that:1) A face recognition system based on this cognition algorithmtrains each type of samples respectively, therefore, a new type ofsamples will not affect the trained ones.2) The training samples in the experiments are continuous, which isthe requirement for a face recognition system based onBiomimetic Pattern Recognition.3) Method of covering of complicated geometrical shapes in high-dimension space is adopted to construct the sample space inBiomimetic Pattern Recognition.The innovative thoughts of this paper are listed below:1) A model of multi-degree of freedom neurons is proposed basedon the high-dimension space geometry.2) A cognition algorithm that is combined with the foundationaltheory of Biomimetic Pattern Recognition is developed.3) Experiments of face recognition prove its efficiency.