An improved feature extraction method for individual offline handwritten digit recognition

Offline handwritten digit recognition (OHDR) is considered as one of difficult problems in the field of pattern recognition. Because it is a challenging computational problem mainly due to the vast differences associated with the handwritten patterns of different individuals. In this paper, a novel method of feature extraction is presented based on structural feature for OHDR by simulating the process of human recognizing handwritten digit. Firstly state and state value are introduced, then the steps of how to determine the eigenvalue is explained in detail, last the method is applied in OHDR, and the result show its effectiveness.