论文部分内容阅读
车字除了规范笔顺,有1个常见笔顺和2个少见笔顺。基于265个训练样本字,使用统计方法,建立了车字的笔顺识别方法。样本字的笔顺由书写者报告。在车字上提取11个特征,建立了决策树形式的分类器。使用212个测试样本字评估分类器性能,分类器拒识率34.4%,95%置信区间[28.0%,40.8%],错误率1.9%,95%置信区间[0,3.7%],准确识别率63.7%,95%置信区间[57.2%,70.2%]。
In addition to the standard car stroke strokes, there is a common stroke sequence and 2 rare sequence. Based on 265 training sample words, a statistical method was used to establish the stroke sequence recognition method of vehicle characters. The stroke order of the sample words is reported by the writer. Eleven features were extracted from the car characters, and a classifier of decision tree was established. 212 test sample words were used to evaluate the classifier performance. The classifier rejection rate was 34.4%, the 95% confidence interval [28.0%, 40.8%], the error rate 1.9%, the 95% confidence interval [0, 3.7%], 63.7%, 95% confidence interval [57.2%, 70.2%].