基于改进的MDF特征的手写体数字识别

Handwritten Numeral Recognition Based on Modified MDF

方向特征是目前手写体识别中最常用和有效的特征之一.为了减少方向值提取过程中带来的误差,对改进的方向特征(MDF)提出了进一步的改进(MMDF),在方向值提取过程中对方向突变条件进行调整,同时引入半方向归一化线段方向并用二维数组来表示方向值.实验证明采用BP神经网络分类器对手写数字进行识别,与MDF相比,MMDF能同时降低拒识率和提高识别精度.

Direction feature is the most commen and effective feature extraction technology in handwritten character recogniton. In or- der to reduce errors during direction value extracting, a number of modifications are proposed to the modifed direction feature （ MDF ） ~ In direction feature value extraction process, some adjustment is performed to the conditon of dirction mutation and half dirction is ~troduction to normalized segment. A two-dimensional array is used to represent the direction value instead of the original one-dimen- sional. The modified MDF{ MMDF） is tested in handwritten numeral database using BP neural network-based classifier and compare to the MDF. MMDF outperformed MDF both in rejection rate and recognition accuray.