融合KPCA与改进KNN的蒙古族音乐分类方法  

Mongolian Music Classification Method Combining KPCA and Improved KNN

作　　者：宋扬 王海龙 柳林 裴冬梅 SONG Yang;WANG Hailong;LIU Lin;PEI Dongmei(College of Computer Science and Technology,Inner Mongolian Normal University,Hohhot,Inner Mongolia 010022,China)

机构地区：[1]内蒙古师范大学计算机科学技术学院,内蒙古呼和浩特010022

出　　处：《复旦学报（自然科学版）》2022年第5期573-580,588,共9页Journal of Fudan University：Natural Science

基　　金：国家重点研发计划(2020YFC1523300);国家级新工科研究与实践项目(E-DSJ20201107);内蒙古自治区自然科学基金(2020MS06030,2021MS06025);内蒙古自治区关键技术应用研究项目(2021GG0184,2021GG0426)。

摘　　要：蒙古族音乐存在传承方式较为落后导致发展缓慢的问题,如何有效利用音乐分类技术对蒙古族音乐进行分类研究,有助于蒙古族音乐的传承与发展。现有的音乐分类方法在特征提取过程中由于特征数量的增多易存在无关和冗余特征,无关和冗余特征反而会带来过拟合的风险,导致分类模型的性能差。针对此问题,本文提出一种融合核主成分分析和改进KNN的分类方法,即KPCA-FSKNN。该方法可以将提取到的特征进行有效降维,避免特征过度冗余,选择出有助于提升分类效果的最佳特征子集。实验结果表明,该方法优于其他分类算法,对蒙古族音乐分类的研究具有重要意义。Mongolian music has the problem of slow development due to the backwardness of its inheritance,so how to effectively use music classification techniques to classify Mongolian music can help the inheritance and development of Mongolian music.The existing music classification methods are prone to irrelevant and redundant features due to the increase in the number of features,and the irrelevant and redundant features may bring the risk of overfitting and lead to poor performance of the classification model.To address this problem,the paper proposes a classification method that combines Kernel Principal Components Analysis(KPCA)and improved KNN,namely KPCA-FSKNN(Kernel Principal Components Analysis-Feature Selection K-Nearest Neighbor).The method can effectively reduce the dimensionality of the extracted features,avoid excessive redundancy of features and select the best subset of features that can help improve the classification effect.The experimental results show that this method outperforms other classification algorithms and is important for the study of Mongolian music classification.

关 键 词：蒙古族音乐 音乐分类 核主成分分析 改进KNN算法 

分 类 号：TP399[自动化与计算机技术—计算机应用技术]