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%。
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
Abstract: 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%.
Key words:
numbered-notation score,
image segmentation,
projection,
lyrics extraction,
Bounding Box(BB) algorithm,
morphology operation,
sliding window
中图分类号:
张立印, 陈根方, 杜旋, 胡海鸥, 赵琼迪. 中文简谱的图像分割与歌词提取研究[J]. 计算机工程, 2012, 38(9): 217-219,236.
ZHANG Li-Yi, CHEN Gen-Fang, DU Xuan, HU Hai-Ou, DIAO Qiong-Di. Research on Image Segmentation and Lyrics Extraction of Chinese Numbered-notation Score[J]. Computer Engineering, 2012, 38(9): 217-219,236.