基于音频特征参数的多语种分类算法  

Multi-Language Classification Algorithm based on Audio Feature Parameters

作　　者：周金傲 龙华[1] ZHOU Jin-ao;LONG Hua(Kunming University of Science and Technology,Kunming Yunnan 650000,China)

机构地区：[1]昆明理工大学,云南昆明650000

出　　处：《通信技术》2018年第10期2350-2355,共6页Communications Technology

摘　　要：伴随着国际化的趋势,音频语种识别问题越来越受到重视。但是,现有的语种识别系统不能满足现代化日益增长的需求。处理小语种和混淆度高的语种分类时,语种识别具有局限性,多语种分类算法的精度和识别率也不理想。针对多语种分类算法精度不高的情况,提出了一种基于改进音频特征参数的多语种分类算法。利用支持向量机作为分类器,设计了多特征语种识别系统,对多语种进行语种识别实验,提取每个语种的特征参数,分别选择单一特征参数集和融合特征参数集输入到SVM分类器中进行测试和训练,分别得到多语种分类的识别率。实验结果证明,使用改进的多语种分类算法可以有效提高多语种分类的精度。With the trend more attention. However, of modernization. When d of internationalization, the issue of audio language recognition attracts more and existing language recognition systems are unable to meet the growing demands ealing with small language and high-degree confusion, language recognition has limitations, and the accuracy and recognition rate of multilingual classification algorithms are also not ideal. Aiming at the low precision of multilingual classification algorithm, a multilingual classification algorithm based on improved audio feature parameters is proposed. With the support vector machine as the classifier, a multi-feature language recognition system is designed, thus to perform language recognition experiments on multiple languages and extract the feature parameters of each language. At the same time, a single feature parameter set and a fusion feature parameter set are respectively selected and inputted into the SVM classifier for testing and training, respectively, and the recognition rate of the multilingual classification is obtained respectively. Experimental results indicate that this modified multilingual classification algorithm could effectively improve the accuracy of multilingual classification. e recognition; support vector machine; feature parameter; multilingual

关 键 词：语种识别 支持向量机 特征参数 多语种 

分 类 号：TP391.9[自动化与计算机技术—计算机应用技术]