本站原创
文章编号:10019081(2013)07193804
doi:10.11772/j.issn.10019081.201
为有效提高语音情感识别系统的识别率，研究分析了一种改进型的支持向量机(SVM)算法。该算法首先利用遗传算法对SVM参数惩罚因子和核函数中参数进行优化，然后用优化后的参数进行语音情感的建模与识别。在柏林数据集上进行7种和常用5种情感识别实验，取得了91.03%和96.59%的识别率，在汉语情感数据集上，取得了97.67%的识别率。实验结果表明该算法能够有效识别语音情感。
关键词:支持向量机；语音情感识别；语音信号；参数优化；遗传算法
:A
英文标题
Speech emotion recognition algorithm based on modified SVM
英文作者名
LI Shuling, LIU Rong*, ZHANG Liuqin, LIU Hong
英文地址(
College of Physical Science and Technology, Central China Normal University, Wuhan Hubei 430079, China
英文摘要)
Abstract:
In order to effectively improve the recognition accuracy of the speech emotion recognition system, an improved speech emotion recognition algorithm based on Support Vector Machine (SVM) was proposed. In the proposed algorithm, the SVM parameters, penalty factor and nuclear function parameter, were optimized with genetic algorithm. Furthermore, an emotion recognition model was established with SVM method. The performance of this algorithm was assessed by computer simulations, and 91.03% and 96.59% recognition rates were achieved respectively in sevenemotion recognition experiments and common fiveemotion recognition experiments on the Berlin database. When the Chinese emotional database was used, the rate increased to 97.67%. The obtained results of the simulations demonstrate the validity of the proposed algorithm.
In order to effectively improve recognition accuracy of the speech emotion recognition system, an improved speech emotion recognition algorithm based on SVM is proposed. In the proposed algorithm, the SVM parameters, penalty factor c and nuclear function parameter g, are optimized by genetic algorithm. Furthermore, an emotion recognition model is established withSVM method. The performance of this algorithm is assessed by computer simulations, and achieve 91.03% and 96.59% recognition rate respectively in the seven emotion recognition experiments and other five common emotion experiments which are operated in the Berlin database and when the Chinese emotional database is used, the rate increases to 97.67%. The obtained results of the simulations demonstrate the validity of the proposed algorithm.
英文关键词Key words:
Support Vector Machine (SVM); speech emotion recognition; speech signal; parameter optimization; Genetic Algorithm （GA）
0 引言
语音是人们交流的主要方式，语音信号不仅传递语义信息，同时承载了说话人的情感状态。情感因素的引入能使人机交互变得更加自然和谐。因此，语音信号的情感识别成为近年来智能人机交互领域的研究热点［1］。语音情感识别是让计算机通过语音信号识别说话者的情感状态，最终实现自然、友好、生动的人机交互。目前国内外学者在这方面进行了大量研究。如美国麻省理工学院(Massachusetts Institute of Technology, MIT)媒体实验室研究的情感机器人［2］，IBM公司的“蓝眼计划”以及美国卡内基梅隆大学(Carnegie Mellon University, CMU)可穿戴计算机的研究开发，这些研究都为情感计算提供了一个较好的研究平台。国内的高校如哈尔滨工业大学、中国科学院计算技术研究所以及中国科学院自动化研究所等，也都在进行人机交互、情感识别方面的研究。
语音信号的情感识别方法很多，常用的情感分析方法有混合高斯分布模型(Gaussian Mixture Model英文全称与中文不匹配，是否正确，请核实。, GMM)法、隐马尔可夫模型(Hidden Markov Model, HMM)、人工神经网络(Artificial Neural Network, ANN)、支持向量机(Support Vector Machine, SVM)以及在这几种识别方法上的改进和组合［3］。Vlasenko等［4］在柏林数据集上采用GMM进行语音情感的识别，取
doi:10.11772/j.issn.1001908
1.2013.07.1938
摘 要:
为有效提高语音情感识别系统的识别率，研究分析了一种改进型的支持向量机(SVM)算法。该算法首先利用遗传算法对SVM参数惩罚因子和核函数中参数进行优化，然后用优化后的参数进行语音情感的建模与识别。在柏林数据集上进行7种和常用5种情感识别实验，取得了91.03%和96.59%的识别率，在汉语情感数据集上，取得了97.67%的识别率。实验结果表明该算法能够有效识别语音情感。关键词:支持向量机；语音情感识别；语音信号；参数优化；遗传算法
:A
英文标题
Speech emotion recognition algorithm based on modified SVM
英文作者名
LI Shuling, LIU Rong*, ZHANG Liuqin, LIU Hong
英文地址(
College of Physical Science and Technology, Central China Normal University, Wuhan Hubei 430079, China
英文摘要)
Abstract:
In order to effectively improve the recognition accuracy of the speech emotion recognition system, an improved speech emotion recognition algorithm based on Support Vector Machine (SVM) was proposed. In the proposed algorithm, the SVM parameters, penalty factor and nuclear function parameter, were optimized with genetic algorithm. Furthermore, an emotion recognition model was established with SVM method. The performance of this algorithm was assessed by computer simulations, and 91.03% and 96.59% recognition rates were achieved respectively in sevenemotion recognition experiments and common fiveemotion recognition experiments on the Berlin database. When the Chinese emotional database was used, the rate increased to 97.67%. The obtained results of the simulations demonstrate the validity of the proposed algorithm.
In order to effectively improve recognition accuracy of the speech emotion recognition system, an improved speech emotion recognition algorithm based on SVM is proposed. In the proposed algorithm, the SVM parameters, penalty factor c and nuclear function parameter g, are optimized by genetic algorithm. Furthermore, an emotion recognition model is established withSVM method. The performance of this algorithm is assessed by computer simulations, and achieve 91.03% and 96.59% recognition rate respectively in the seven emotion recognition experiments and other five common emotion experiments which are operated in the Berlin database and when the Chinese emotional database is used, the rate increases to 97.67%. The obtained results of the simulations demonstrate the validity of the proposed algorithm.
英文关键词Key words:
Support Vector Machine (SVM); speech emotion recognition; speech signal; parameter optimization; Genetic Algorithm （GA）
0 引言
语音是人们交流的主要方式，语音信号不仅传递语义信息，同时承载了说话人的情感状态。情感因素的引入能使人机交互变得更加自然和谐。因此，语音信号的情感识别成为近年来智能人机交互领域的研究热点［1］。语音情感识别是让计算机通过语音信号识别说话者的情感状态，最终实现自然、友好、生动的人机交互。目前国内外学者在这方面进行了大量研究。如美国麻省理工学院(Massachusetts Institute of Technology, MIT)媒体实验室研究的情感机器人［2］，IBM公司的“蓝眼计划”以及美国卡内基梅隆大学(Carnegie Mellon University, CMU)可穿戴计算机的研究开发，这些研究都为情感计算提供了一个较好的研究平台。国内的高校如哈尔滨工业大学、中国科学院计算技术研究所以及中国科学院自动化研究所等，也都在进行人机交互、情感识别方面的研究。
语音信号的情感识别方法很多，常用的情感分析方法有混合高斯分布模型(Gaussian Mixture Model英文全称与中文不匹配，是否正确，请核实。, GMM)法、隐马尔可夫模型(Hidden Markov Model, HMM)、人工神经网络(Artificial Neural Network, ANN)、支持向量机(Support Vector Machine, SVM)以及在这几种识别方法上的改进和组合［3］。Vlasenko等［4］在柏林数据集上采用GMM进行语音情感的识别，取
源于：毕业论文致谢词范文www.udooo.com
得了89.9%的识别率。但GMM需要大量的情感语音样本，且在情感模型的训练上花费时间较长。文献［5］采用改进的蛙跳算法(Shuffled Frog Leaping Algorithm, LA)神经网络进行语音情感的识别研究，得到了84.2%的识别率。但ANN中隐藏节点如何选取不确定，并且隐藏节点数目越多，网络的结构就会越复杂。文献［6］结合HMM和ANN两种算法在自制语料库上识别6种语音情感(平静、高兴、惊奇、愤怒、悲伤、恐惧)，其识别率在69.1%~94.8%。文献［7］进行基于SVM的语音情感识别，在柏林数据集上获得了86.36%的识别率。SVM情感分析方法在解决非线性、小样本以及高维模式识别表现出特有的优势，因而受到广泛的关注。但SVM核函数及其参数的选择，目前国际上还没有统一标准，一般是多次尝试取其经验值［8］。