第13卷第2期 智能系统学报 Vol.13 No.2 2018年4月 CAAI Transactions on Intelligent Systems Apr.2018 D0:10.11992/tis.201612012 网络出版地址：http:/kns.cnki.net/kcms/detail/23.1538.TP.20170508.0922.002.html 基于测度学习支持向量机的钢琴乐谱难度等级识别 郭龙伟，关欣，李锵 (天津大学电子信息工程学院，天津300072) 摘要：现有钢琴乐谱难度分类主要由人工方式完成.效率不高，而自动识别乐谱难度等级的算法对类别的拟合度较 低。因此，与传统将乐谱难度等级识别归结为回归问题不同，本文直接将其建模为基于支持向量机的分类问题。并 结合钢琴乐谱分类主观性强、特征之间普遍存在相关性等特点，利用测度学习理论有难度等级标签乐谱的先验知识 依据特征对难度区分的贡献度，改进高斯径向基核函数，从而提出一种测度学习支持向量机分类算法一一ML SVM算法。在9类和4类难度两个乐谱数据集上，我们将ML-SVM算法与逻辑回归，基于线性核函数、多项式核函 数、高斯径向基核函数的支持向量机算法以及结合主成分分析的各个支持向量机算法进行了对比，实验结果表明我 们提出算法的识别正确率优于现有算法，分别为68.74%和84.67%。所提算法有效提高了基于高斯径向基核函数支 持向量机算法在本应用问题中的分类性能。 关键词：数字钢琴乐谱：难度等级识别：分类算法；支持向量机：测度学习；高斯径向基核函数 中图分类号：TP391.4文献标志码：A文章编号：1673-4785(2018)02-0196-06 中文引用格式：郭龙伟，关欣，李锵.基于测度学习支持向量机的钢琴乐谱难度等级识别.智能系统学报，2018,13(2：196-201. 英文引用格式：GUO Longwei,,GUAN Xin,LI Qiang.Recognition of difficulty level of piano score based on metric learning sup- port vector machineJ.CAAI transactions on intelligent systems,2018,13(2):196-201. Recognition of difficulty level of piano score based on metric learning support vector machine GUO Longwei,GUAN Xin,LI Qiang (Department of Electronic Information Engineering.Tianjin University,Tianjin 300072,China) Abstract:The existing classification work about piano score's level is mainly done manually and inefficient,while the algorithm automatically recognizing the difficulty class of music scopre has a low classification fitting degree.There- fore,different from the traditional method that takes the recognition for the difficulty class of music scope as a regres- sion issue,the paper directly modelled it as a classification based on the support vector machine,in addition,in combin- ation with such characteristics of the score classification as intense subjectivity and common dependency among fea- tures,the metric learning theory was utilized.The prior knowledge of the score with difficult level tag was sufficiently utilized,according to the contribution of feature in difficulty distinguishment,the Gauss radial basis kernel function was improved,so as to propose a kind of metric learning support vector machine classification algorithm-ML-SVM al- gorithm.In the score datasets with level 9 and level 4 difficulty,ML-SVM algorithm was compared with logistic regres- sion,the support vector machine algorithm based on linear kernel function,polynomial kernel function,Gauss radical basis(GRB)kernel function,and various support vector machine algorithms combining principal component analysis. The results show that the proposed algorithm is much more accurate than the existing algorithms,reaching the accuracy rate 68.74%and 84.67%respectively.The proposed algorithm effectively improves the classification performance of SVM algorithm based on GRB kernel function in this application. Keywords:digital piano score;recognition of difficulty level;classification algorithm;support vector machine;metric learning;Gauss radial basis kernel function 收稿日期：2016-12-08.网络出版日期：2017-05-08 虽然互联网中存在海量的钢琴乐谱资源，但其 基金项目：国家自然科学基金项目(60802049,61471263)：天津市 自然科学基金重点项目(16 JCZDJC31100). 难易程度不一，对于业余和初级乐器学习者来说， 通信作者：关欣.E-mail:guanxin@tju.edu.cn. 由于缺乏专业知识和指导，难以有效地找到与自身DOI: 10.11992/tis.201612012 网络出版地址: http://kns.cnki.net/kcms/detail/23.1538.TP.20170508.0922.002.html 基于测度学习支持向量机的钢琴乐谱难度等级识别 郭龙伟，关欣，李锵 （天津大学 电子信息工程学院，天津 300072） 摘 要：现有钢琴乐谱难度分类主要由人工方式完成，效率不高，而自动识别乐谱难度等级的算法对类别的拟合度较 低。因此，与传统将乐谱难度等级识别归结为回归问题不同，本文直接将其建模为基于支持向量机的分类问题。并 结合钢琴乐谱分类主观性强、特征之间普遍存在相关性等特点，利用测度学习理论有难度等级标签乐谱的先验知识， 依据特征对难度区分的贡献度，改进高斯径向基核函数，从而提出一种测度学习支持向量机分类算法——ML￾SVM 算法。在 9 类和 4 类难度两个乐谱数据集上，我们将 ML-SVM 算法与逻辑回归，基于线性核函数、多项式核函 数、高斯径向基核函数的支持向量机算法以及结合主成分分析的各个支持向量机算法进行了对比，实验结果表明我 们提出算法的识别正确率优于现有算法，分别为 68.74% 和 84.67%。所提算法有效提高了基于高斯径向基核函数支 持向量机算法在本应用问题中的分类性能。 关键词：数字钢琴乐谱；难度等级识别；分类算法；支持向量机；测度学习；高斯径向基核函数 中图分类号：TP391.4 文献标志码：A 文章编号：1673−4785(2018)02−0196−06 中文引用格式：郭龙伟, 关欣, 李锵. 基于测度学习支持向量机的钢琴乐谱难度等级识别[J]. 智能系统学报, 2018, 13(2): 196–201. 英文引用格式：GUO Longwei, GUAN Xin, LI Qiang. Recognition of difficulty level of piano score based on metric learning sup￾port vector machine[J]. CAAI transactions on intelligent systems, 2018, 13(2): 196–201. Recognition of difficulty level of piano score based on metric learning support vector machine GUO Longwei，GUAN Xin，LI Qiang (Department of Electronic Information Engineering, Tianjin University, Tianjin 300072, China) Abstract: The existing classification work about piano score’s level is mainly done manually and inefficient, while the algorithm automatically recognizing the difficulty class of music scopre has a low classification fitting degree. There￾fore, different from the traditional method that takes the recognition for the difficulty class of music scope as a regres￾sion issue, the paper directly modelled it as a classification based on the support vector machine, in addition, in combin￾ation with such characteristics of the score classification as intense subjectivity and common dependency among fea￾tures, the metric learning theory was utilized. The prior knowledge of the score with difficult level tag was sufficiently utilized, according to the contribution of feature in difficulty distinguishment, the Gauss radial basis kernel function was improved, so as to propose a kind of metric learning support vector machine classification algorithm —ML-SVM al￾gorithm. In the score datasets with level 9 and level 4 difficulty, ML-SVM algorithm was compared with logistic regres￾sion, the support vector machine algorithm based on linear kernel function, polynomial kernel function, Gauss radical basis (GRB) kernel function, and various support vector machine algorithms combining principal component analysis. The results show that the proposed algorithm is much more accurate than the existing algorithms, reaching the accuracy rate 68.74% and 84.67% respectively. The proposed algorithm effectively improves the classification performance of SVM algorithm based on GRB kernel function in this application. Keywords: digital piano score; recognition of difficulty level; classification algorithm; support vector machine; metric learning; Gauss radial basis kernel function 虽然互联网中存在海量的钢琴乐谱资源，但其 难易程度不一，对于业余和初级乐器学习者来说， 由于缺乏专业知识和指导，难以有效地找到与自身 收稿日期：2016−12−08. 网络出版日期：2017−05−08. 基金项目：国家自然科学基金项目 (60802049，61471263)；天津市 自然科学基金重点项目 (16JCZDJC31100). 通信作者：关欣. E-mail：guanxin@tju.edu.cn. 第 13 卷第 2 期 智 能 系 统 学 报 Vol.13 No.2 2018 年 4 月 CAAI Transactions on Intelligent Systems Apr. 2018