摘要
在音乐旋律特征相似度匹配理论和DTW算法基础上,提出了一种基于哼唱片段的相似度匹配算法,减少了对歌曲库进行精确匹配(DTW)的歌曲匹配数量并利用不同匹配路径进行精确匹配的性能进行分析和对于基于内容的音乐精确匹配算法,提出了用于加快检索速度的加速方法。设计实现了一个基于内容的音乐哼唱检索系统。经过实验验证该系统对于小规模数据库哼唱歌曲片段的检索具有较高的准确性,对进一步的深入研究具有推动和借鉴意义。
Through analyzing the theoreties of similarity comparison and DTW algorithm,A similarity comparison algorithm of content-base query by humming segment is proposed,it reduce the number of precision comparison . Analyzing the capability of precision comparison in different path.and A faster match of content-base query of precison comparison is proposed,it reduces the time in retrieving. A content-base query by humming system is designed and completed .The experimental results show that the system has high recognition rate and pushes forward the study on melody retrieval.
出处
《信息与电脑（理论版）》
2009年第12期25-26,共2页
China Computer & Communication
关键词
哼唱识别
DTW
旋律特征
相似度匹配
humming identify
DTW
melody features
similarity comparison.
