基于DTW算法的哼唱检索系统的设计

本文在音乐主旋律特征的提取、特征库的建立方法以及以DTW算法为核心的相似度匹配算法的理论原理基础之上，提出一种针对MIDI音乐的主旋律提取方法。该方法简化了提取音乐特征的工作量，缩短了检索时间。设计实现了一个基于音乐内容的哼唱检索系统，利用包含200首乐曲特征的数据库对该系统进行验证。实验结果表明，该系统具有较高的检索命中率和较快的检索速度。

Query By Humming System Design Base On the DTW Algorithm

This article analyzed the music melody feature extraction, the method of build feature library and the theories of similarity comparison which was based on the DTW algorithm. A music melody feature extraction that suitable for MID1 files is proposed. The method simplified the workload to extract music meh）dy feature and reduce the retrieval time. A content-base query by humming system is designed and completed. The experimental results show that the system has high recognition rate and fast retrieval speed.