中文简谱的图像分割与歌词提取研究  

Research on Image Segmentation and Lyrics Extraction of Chinese Numbered-notation Score

作　　者：张立印[1] 陈根方[1] 杜旋[1] 胡海鸥[1] 赵琼迪[1] 

机构地区：[1]杭州师范大学信息科学与工程学院,杭州310036

出　　处：《计算机工程》2012年第9期217-219,236,共4页Computer Engineering

基　　金：浙江省大学生科技创新活动计划基金资助项目(2010R421055);杭州市属高校重点实验室科技创新基金资助项目(20100331T18);杭州师范大学硕士研究生创新基金资助项目

摘　　要：为准确获取简谱的歌词信息,提出一种中文简谱的图像分割与歌词提取方案。在图像分割阶段,利用投影法与中文歌谱的高层领域知识分离歌词与乐音、标题等区域;在歌词提取阶段,设计包围盒算法提取歌词,合并属于同一个字的包围盒,区分标点符号,并与歌词包围盒分开存储。实验结果表明,该方案具有较好的歌词提取效果,平均歌词提取率可达97.48%。In order to acquire lyrics of numbered-notation score, this paper proposes an scheme of image segmentation and lyrics extraction for Chinese numbered-notation score. During the image segmentation stage, it separates lyrics from music and title regions utilizing the projection and high-level domain knowledge of numbered-notation score. During the lyrics extraction stage, it designs Bounding Box（BB） algorithm to extract lyrics, merges the boxes belonging to one character, distinguishes punctuations from lyrics and stores them separately. Experimental results show that this scheme can extract lyrics effectively, and the average rate of lyrics extraction reaches 97.48%.

关 键 词：简谱 图像分割 投影 歌词提取 包围盒算法 形态学运算 滑动窗口 

分 类 号：TN911.73[电子电信—通信与信息系统]