论文部分内容阅读
本文提出了一种可在8位单片机上实现的非特定人汉语数码识别算法。该算法基于最大后验概率分类原理，利用二次多项式估计后验概率分布。其优点是识别算法简单，不涉及动态规划过程。又针对易混淆的集合设计了包括优选、拒识策略的两级结构，实验证明本算法达到了实用水平。
This paper presents a non-specific Chinese digital recognition algorithm that can be implemented on an 8-bit microcontroller. Based on the principle of maximum a posteriori probability classification, the algorithm uses quadratic polynomial to estimate the posterior probability distribution. The advantage is that the recognition algorithm is simple and does not involve the dynamic planning process. In addition, a two-level structure including preference and rejection strategies is designed for the confusable set. Experiments show that the algorithm reaches a practical level.