切线距离的SVD方法在汉字识别细分问题中的研究

Study on tangent distance with SVD method for characters recognition fine classification

针对手写体汉字字形变化复杂,而传统的欧式距离对汉字形变比较敏感。将基于切线距离的SVD分解选取切线向量方法应用于大字符集脱机手写体汉字识别细分类中,取得了较好的识别结果。实验证明,采用SVD选取切线向量的方法具有较好的通用性和重复性,适合应用于手写体汉字识别。

The tangent distance with SVD method was applied to large-set offline hand-written Chinese characters＇ fine classification since the traditional Euclidean distance is sensitive to the hand-written characters＇ shape transformation. Experiment indicates that this method has better recognition accuracy and proves the efficiency of the tangent distance classifier.