Handwritten Digit Recognition Based on LDA and AP
Due to the instability of handwritten character pattern caused by different writing styles, excessively enormous training set is often used to obtain an outstanding recognition rate. To overcome the problem caused by enormous training set, a novel approach based on AP and LDA is proposed. Effective discriminate features are extracted, accordingly, the new training set which is stable and effectual are constructed. The main contribution of the algorithm proposed in this paper is to improve the training set and enhance the efficiency of recognition obviously retaining high recognition accuracy and reliability simultaneously. Experiments have been performed with CEXPARMI, Mnist and NUST handwritten digit database, and the results demonstrate the robustness and effectiveness of this algorithm..
Handwritten digit recognition Affinity propagation Linear discriminant analysis representative pattern detection
Yan Zhang Jian Gu
The Third Research Institute of Ministry of Public Security Shanghai, China
国际会议
太原
英文
144-147
2011-02-26（万方平台首次上网日期，不代表论文的发表时间）