摘要： 中国传统戏曲是世界艺术园地中的一支奇葩。该文使用音频分析技术和模式识别技术相结合的方法对8 种典型的中国传统戏曲（京剧、昆剧、评剧、豫剧、越剧、河北梆子、黄梅戏和晋剧）进行了自动分类和相似性分析研究。实验结果表明，在一个包含了680 个唱段的测试数据库上，该方法可以达到82.4％的平均分类正确率。研究结果还表明在这8 种传统戏曲中，昆剧和评剧最为相似，豫剧和越剧之间差别最大。

关键词: 戏曲；音频分析；模式识别

Abstract: Among all kinds of music styles, Chinese traditional opera is very familiar in China. This paper presents a study on content-based classification and analysis among eight traditional opera styles (including Beijing opera, Kunqu opera, Pingju opera, Henan opera, Shaoxing opera, Hebei opera, Huangmei opera and Jin opera), using audio analysis techniques together with pattern recognition techniques. A comparative evaluation between different classifiers is carried out on a test database of 680 audio segments, the result show that the quadratic classifier (QC) works best, and its average classification accuracy can achieve 82.4%. By analyzing the features distribution and the classification results of these traditional opera styles, some interesting conclusions are also reported

Key words: Chinese traditional opera; Audio analysis; Pattern recognition