基于改进的T型关联度在音乐分类中的应用

发布时间：2018-03-14 01:37

  本文选题：灰色T型关联度　切入点：改进　出处：《数学的实践与认识》2017年20期 　论文类型：期刊论文

【摘要】：在音乐分类问题中,绝大多数的算法需要提取多个特征值进行分析,工作量和复杂程度也随之增加,并且分类太绝对化.为了降低工作量和复杂程度,采用新的方式对音乐进行分类,即引进灰色关联度分析方法.现有的灰色T型关联度模型均存在不足,对序列采用绝对初值化处理,并且对关联系数计算公式进行改进,增强了结果的准确性和可信度.将提取出的短时能量、短时平均过零率和短时平均幅度作为音乐分类的三大特征值,对大部分音乐进行了较为准确的分类,排除率达到90.1%.而且此方法能够体现出各音乐之间的关联程度,使得分类更加人性化,这点具有现实意义.并且避免了复杂的计算过程和巨大的工作量,简化了解决问题的方式,也减少了对特征值的依赖,仅采用三种特征就达到了很好的效果.这充分反映了思路的正确性、实用性和可行性.
[Abstract]:In the music classification problem, most of the algorithms need to extract multiple eigenvalues for analysis, the workload and complexity also increase, and the classification is too absolute. In order to reduce the workload and complexity, The music is classified in a new way, that is, the grey correlation degree analysis method is introduced. The existing grey T-type correlation degree models are all deficient, the sequence is treated with absolute initial value, and the formula for calculating the correlation coefficient is improved. The accuracy and reliability of the results are enhanced. The extracted short time energy, short time average zero crossing rate and short time average amplitude are taken as the three characteristic values of music classification, and the majority of music is classified more accurately. The exclusion rate is 90.1 and this method can reflect the degree of relevance between the various music, which makes the classification more humanized, and avoids the complicated calculation process and huge workload, and simplifies the way to solve the problem. It also reduces the dependence on the eigenvalues and achieves good results by using only three features, which fully reflects the correctness, practicability and feasibility of the idea.
【作者单位】： 海南大学信息科学技术学院;
【基金】：海南省自然科学基金(117011) 海南省教育厅高校科研资助项目(Hnky2017-12) 海南大学青年基金(hdkyxj201719);海南大学教育教学科研资助项目(hdjy1639)
【分类号】：J61;N941.5
，

本文编号：1609034