中国民歌地域风格分类中的特征选择.PDF

12A Vol. 36 No. 12A 2008 12 ACTA ELECTRONICA SINICA Dec. 2008 刘 怡, 蔚 磊, 刘子利, 王 鹏 ( , 100872) : , . . 10 1392 , . : , SVM ; SVM Active Feature Selection , 83% , 74 3 , . : ; ; ; : TP37, TP181 : A : (2008) 12A1 20 The Feature Selection of Regional Style Classification of Chinese Folk Songs LIU Yi, WEI Lei, LIU Zili,WANG Peng ( Inf ormation School, Renmin Universi ty of China , Beji ing 100872, China) Abstract : Automatic music style classification is an important topic in the area of music information retrieval. In this paper, we present a study on automatic classification of Chinese folk songs by regional style which mainly focuses on performance of dif ferent feature selection method. We did music style classification experiments on 1392 original Chinese folk songswhich are collect ed from 10 different regions.The experiment results show that support vector machine classification performance has a certain ad vantage among different classifierswithout feature selection. Simultaneously, support vector machine get the highest classification ac curacy of 83% with active feature selection method, the feature vector dimensions are reduced from 74 to 3 using active feature se lection feature selection.Therefore, the selected feature set is more convenient for music analysis. Key words: music information retrieval; music data mining; music style feature selection 1 , 11 , , . , , , . , . , ,