基于DTW的哼唱检索系统分析
首发时间:2012-02-21
摘要:哼唱检索系统中匹配算法模块,决定系统的检索准确率和检索时间,因此匹配算法一直是哼唱检索的热点研究问题。动态时间规整(DTW)算法作为一种非线性匹配算法,在哼唱检索系统中得到广泛的应用。本文通过在DTW算法中引入偏移量,优化DTW算法提升哼唱检索系统的性能。同时考虑到用户哼唱音乐片段时节奏快慢的变化,本文通过对输入音乐片段做线性伸缩处理,降低因前段输入引入的误差。最后结合试验结果给出系统性能的分析。
For information in English, please click here
Analyse of Query by Humming System based on Dynamic Timing Warping
Abstract:In Query By Humming System, the matching algorithm determine the retrieval accuracy and retrieval time, so the matching algorithm is always the hot research problem. Dynamic Timing Warping (DTW) algorithms as a nonlinear matching algorithm, which is widely used in Query By Humming System. This paper introduced the DTW algorithm offset to optimize the System performance. Taking into account the rhythm of the pace of change in user humming the music clips, this paper introduced the linear scaling of the input music clips to reduce the preceding error. Finally, based on the test result to show the System performance analysis.
Keywords: DTW Query by Humming Linear Scaling Pitch Frequencies
基金:
论文图表:
引用
No.****
同行评议
共计0人参与
勘误表
基于DTW的哼唱检索系统分析
评论
全部评论