Geometric algebra multi-vector representation method of pattern features
Zhi Yang Gao · Journal of Yanshan University · 2010
Pattern representation is a basic problem of pattern recognition. In traditional statistical pattern recognition theory, pattern features are usually represented as a numeric vector and can be considered as a point in an dimensional Euclidean space. This representation model uses only one order features, is prone to lose the interrelation of multiple features and higher order structure. Firstly the axiom definition of geometric algebra and some basic concepts is introduced, then the traditional pattern vector representation is generalized to the multi-vector representation in geometric algebra space. Two special cases of this representation are discussed. The basic framework of pattern recognition based on the multi-vector representation is presented. In conclusion, the prospect of geometric algebra applying to visual pattern recognition and work to do in the future are outlined.