基于隐式马尔可夫模型的遗传类比学习在中国书法生成中的应用

基于隐式马尔可夫模型的遗传类比学习在中国书法生成中的应用 Chinese Calligraphy Production Method Based on HMM Genetics Analogy 徐杨 XU Yang 1 first-author 中南民族大学工商学院,湖北,武汉,430223 中南民族大学工商学院,湖北,武汉,430223 Engineering and Commerce College, South-Central University for Nationalities Engineering and Commerce College, South-Central University for Nationalities 针对非物质文化遗产虚拟修复中缺失的书法字体生成问题,实现了一种基于隐式马尔可夫模型（HMM）的曲线类比学习书法生成系统.该系统首先收集多种标准书法字体以及骨架作为学习样本,然后运用有指导的类比学习,样本字库中的字符骨架在遗传演化的过程中形成风格上与待修复字体风格接近的骨架结构,最后用HMM对骨架进行各种丰满度处理以获得连笔、枯笔等多种中国书法特有的形态特征.实验证明本文的算法能产生与原作品风格趋近一致的书法汉字并获取新颖的字体风格. A calligraphy production system based on HMM （hidden Markov model） curve analogy learning is introduced for the reparation of the lost calligraphy character, which can make some kinds of calligraphies according to the environmental style. The system collects various calligraphy characters and skeletons styles. A fter the genetic evolution algorithms proceeding, the curve analogy learning function transforms the old skeletons to new ones,which are similar to the style of the lost calligraphy character. At last, the skeletons are processed to fulfill the unique calligraphy styles,and the complete character is produced. The system can produce the similar style of the lost calligraphy characters and some experimental results are novel. 曲线类比 字符骨架化 隐式马尔可夫模型 用户满意度参数 欧几里德距离 curve analogy character skeleton extraction HMM（hidden Markov model） users’ content parameter Euclid distance TP18 国家自然科学基金资助项目（60672051） 2008-01-01 2021-04-01 1