由于手写哈萨克字符结构的特殊性,仅提取几种单一的字符特征进行识别时正确率较低,识别效果较差.由此采用改进的PCA方法定位单词基线位置,对每个字符提取包括笔画密度特征、投影特征、轮廓特征等在内的36种特征,使用K-W检验对各特征的分类能力进行比较,并采用线性判别函数进行分类,取得了较高的识别精度.实验结果表明,该系统针对脱机字符识别率达到94%以上.%Owing to the special structure of handwritten Kazakh characters, just using several simple features to recognize Kazakh characters encounter the problems of low recognition rate and poor recognition effects. This paper determines the word baseline by using improved PCA method, extracts 36 features for each character and the K-W checking is used to compare the classification ability of each feature. It uses the linear discriminant function to classify the characters and acquires a higher recognition rate. Experimental result shows that to off-line handwritten character the recognition accuracy achieves 94％.
展开▼
