Image Recognition Based on High-Dimensional Information Geometry
Peng Chen, Ning Liu · Applied Mechanics and Materials · 2015
This paper proposes a new method of image recognition based on the theory of high-dimensional information geometry. Firstly, we used the linear algebra to describe the high-dimensional information geometry and studied the effectiveness of included angle cosine in different dimensional spaces. Secondly, the image could be expressed as a vector and the cosine value of angle between two vectors was used as the discriminant conditions of classifier to identify images. Finally, a face recognition system extracting facial features based on a fast principal component analysis (PCA) was designed, using Euclidean distance and included angle cosine value to measure the nearest neighbor classifier respectively for comparision. The experiments results proved that the proposed method has a higher accuracy and better performance than a traditional recognition system.