基于多分辨率直方图的哼唱查询渐进过滤
Progressive Filtering Using Multiresolution Histograms for Query by Humming System
摘要:
数字音乐的日益普及要求有效的分类和检索方法。真实世界的场景的特征在于通过参考用户输入的相关和非相关歌曲的庞大的音乐集合。研究工作的主要目标是通过渐进过滤(PF)哼唱查询(QBH)系统来平衡不相关歌曲的危险影响。PF是一种通过缩减空间解决问题的技术。提出了基于多分辨率直方图(MRH)的粒子滤波器(PF)的概念,并给出了其有效的设计方法,以实现流形上的搜索。首先搜索整个音乐数据库,以获得高召回率和缩小的搜索空间。后面的步骤在减少的外围完成缓慢的搜索,并实现额外的精度。实验上的大型音乐数据库,使用递归编程证实了该方法的潜力。所提出的策略的结果瞥见,MRH有效地定位模式。在较低水平的MRH的距离是在较高水平的距离的下限,这保证了规避假解雇过程中PF。在适当的时候,所提出的方法有助于在效率和效果之间取得平衡。该系统是可扩展的大型音乐检索系统和数据驱动的性能优化作为一个额外的优势。
英文摘要:
The rising availability of digital music stipulates effective categorization and retrieval methods. Real world scenarios are characterized by mammoth music collections through pertinent and non-pertinent songs with reference to the user input. The primary goal of the research work is to counter balance the perilous impact of non-relevant songs through Progressive Filtering (PF) for Query by Humming (QBH) system. PF is a technique of problem solving through reduced space. This paper presents the concept of PF and its efficient design based on Multi-Resolution Histograms (MRH) to accomplish searching in manifolds. Initially the entire music database is searched to obtain high recall rate and narrowed search space. Later steps accomplish slow search in the reduced periphery and achieve additional accuracy. Experimentation on large music database using recursive programming substantiates the potential of the method. The outcome of proposed strategy glimpses that MRH effectively locate the patterns. Distances of MRH at lower level are the lower bounds of the distances at higher level, which guarantees evasion of false dismissals during PF. In due course, proposed method helps to strike a balance between efficiency and effectiveness. The system is scalable for large music retrieval systems and also data driven for performance optimization as an added advantage.
机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。