基于模式识别方法的果蝇振翅声分类研究
|
摘要
果蝇是一种对水果危害较大的昆虫,它给人类带来巨大的经济损失。大量学者通过研究发现,果蝇在飞行过程中会通过翅膀的振动产生一种振翅声,并以此传递信息,进行种间交流。这种振翅声不仅可以反映果蝇种的特异性,而且还有利于果蝇害虫的诱集、捕杀。所以,对果蝇振翅声的研究具有极大的生物学意义和实践意义。
研究学者们发现果蝇的振翅声中有一种求偶歌,它在果蝇的捕捉、培养等方面都有很大的作用。但是,通过果蝇的振翅声分类果蝇品系和雌雄的相关研究相对较少,不够全面。本文以果蝇的振翅声为研究对象,采用模式识别的方法分类果蝇的品系和雌雄。结果表明,本文方法对于果蝇的品系识别和雌雄分类是可行且有效的,为果蝇的分类研究提供了新的方法。
本文的研究主要包括以下几个方面：
(1)概述了果蝇鸣声的国内外研究状况以及当前研究存在的问题。
(2)介绍了隐马尔可夫模型的识别原理和算法,并综述了隐马尔可夫模型用于动物鸣声的研究现状。
(3)实现了对3个不同品系果蝇振翅声的识别。首先,介绍了梅尔倒谱系数方法以及高斯混合模型和隐马尔可夫模型相结合方法的基本原理；其次,根据本文实验数据选择梅尔倒谱系数的最优参数,提取不同品系果蝇振翅声的有效特征组合,即：Mel倒谱系数的能量特征和一阶方差特征；最后,将高斯混合模型和隐马尔可夫模型结合为每个品系果蝇的振翅声建立模型,根据输出观察值概率的最大值识别未知品系果蝇的振翅声,并计算识别结果。实验结果表明,此方法对于不同品系果蝇的识别是有效的,且效果良好。
(4)实现了对2个品系果蝇振翅声的雌雄性别分类。首先,深入研究了自回归模型和支持向量机的原理；其次,结合本文实验数据选择合适的自回归模型参数和算法,使用AIC准则计算自回归模型的阶数,用Burg算法估计同一品系雌性果蝇和雄性果蝇振翅声的功率谱,提取代表不同性别果蝇振翅声特性的最优特征组合,即：自回归模型的系数和白噪声序列方差；最后根据十折交叉验证方法选择支持向量机分类器的最优参数,采用核函数为重尾径向基函数的支持向量机,通过训练样本集得到最优分类界面,对测试样本集进行分类,并计算分类结果。实验结果表明,此方法对于分类同一品系果蝇的雌雄是可行且有效的,实验效果良好。
Fruit flies are a kind of insects, which is larger harmful to fruits, and they will bring about huge economic losses to human. Through studying, a large number of scholars discover that fruit flies product sounds through their wings' vibrations in flight, and use them to transmit information and communicate among their species. The wings vibration sounds not only can reflect the species-specificity of fruit fly, but also can have advantage to fruit flies pests trapping and killing. Therefore, it has great biological importance and practical significance to research on fruit fly wings vibration sounds.
Scholars discovered courtship song among wings vibration sounds of fruit flies, which has very significant role in fruit flies capture and training. But the related research, which is about the classification of strains and gender through their wings vibration sounds, is opposite less, not comprehensive. So, in the article, wings vibration sounds of fruit flies are used as research object, and pattern recognition methods are tried to implement the classification of strains and gender of fruit flies. According to the experiment, we discover that the article's method is feasible and effective to strains identification and gender classification of fruit flies and provides a new method for fruit flies classification research.
The article's research mainly includes the following aspects:
(1) Summarizing research situations on fruit flies in home and abroad and problems existing in current research.
(2) Introducing the recognition principles and algorithms about Hidden Markov Model, and reviewing the current study status that Hidden Markov Model is used for animal sounds.
(3) Achieving the recognition for wings vibration sounds of three different strains of fruit flies. Firstly, it is introduced that the basic principle of Mel-frequency Cepstrum Coefficients method and the combination method of Gaussian Mixture Model and Hidden Markov Model. Secondly, according to the experimental data in the paper, the optimal parameters are selected for Mel-frequency Cepstrum Coefficients method, the effective features combination of wings vibration sounds of different strains of fruit flies, which is made of the energy feature and first-order variance characteristics, is extracted. Finally, models are established for wings vibration sounds of each strain fruit flies using the combination model with Gaussian Mixture Model and Hidden Markov Model, the wings vibration sound of unknown strain fruit fly is identified according to the max probability value of observation produced, and the recognition results are calculated. Experimental results show that the article's method is effective to the recognition of different strains fruit flies, and the results are good.
(4) Realizing the gender classification between female and male for wings vibration sounds of two strains of fruit flies. In the first, the theory of Auto-Regressive Model and Support Vector Machine are deeply studied. Then, combined with the experimental data, the appropriate parameters and algorithms of Auto-Regressive Model are selected, the orders are calculated by AIC criterion, the power spectrums of wings vibration sounds of female fruit flies and male fruit flies, which are the same strain, are estimated through Burg algorithm, the features combination, which can optimally reflect characteristics of wings vibration sounds of different gender fruit flies, includes the coefficients of Auto-Regressive Model and white noise sequence variance. Finally, the optimal parameters, which are suitable to the classifier of Support Vector Machine, are got by the ten fold cross-validation method, and using the heavy-tailed radial basis function as the kernel function of Support Vector Machine, the optimal classification interface is obtained through training sets, the testing set is classified, and the results are computed. Experimental results show that the article's method is feasible and effective for the classification between female and male in the same strain of fruit fly, and the results are good.
研究学者们发现果蝇的振翅声中有一种求偶歌,它在果蝇的捕捉、培养等方面都有很大的作用。但是,通过果蝇的振翅声分类果蝇品系和雌雄的相关研究相对较少,不够全面。本文以果蝇的振翅声为研究对象,采用模式识别的方法分类果蝇的品系和雌雄。结果表明,本文方法对于果蝇的品系识别和雌雄分类是可行且有效的,为果蝇的分类研究提供了新的方法。
本文的研究主要包括以下几个方面：
(1)概述了果蝇鸣声的国内外研究状况以及当前研究存在的问题。
(2)介绍了隐马尔可夫模型的识别原理和算法,并综述了隐马尔可夫模型用于动物鸣声的研究现状。
(3)实现了对3个不同品系果蝇振翅声的识别。首先,介绍了梅尔倒谱系数方法以及高斯混合模型和隐马尔可夫模型相结合方法的基本原理；其次,根据本文实验数据选择梅尔倒谱系数的最优参数,提取不同品系果蝇振翅声的有效特征组合,即：Mel倒谱系数的能量特征和一阶方差特征；最后,将高斯混合模型和隐马尔可夫模型结合为每个品系果蝇的振翅声建立模型,根据输出观察值概率的最大值识别未知品系果蝇的振翅声,并计算识别结果。实验结果表明,此方法对于不同品系果蝇的识别是有效的,且效果良好。
(4)实现了对2个品系果蝇振翅声的雌雄性别分类。首先,深入研究了自回归模型和支持向量机的原理；其次,结合本文实验数据选择合适的自回归模型参数和算法,使用AIC准则计算自回归模型的阶数,用Burg算法估计同一品系雌性果蝇和雄性果蝇振翅声的功率谱,提取代表不同性别果蝇振翅声特性的最优特征组合,即：自回归模型的系数和白噪声序列方差；最后根据十折交叉验证方法选择支持向量机分类器的最优参数,采用核函数为重尾径向基函数的支持向量机,通过训练样本集得到最优分类界面,对测试样本集进行分类,并计算分类结果。实验结果表明,此方法对于分类同一品系果蝇的雌雄是可行且有效的,实验效果良好。
Fruit flies are a kind of insects, which is larger harmful to fruits, and they will bring about huge economic losses to human. Through studying, a large number of scholars discover that fruit flies product sounds through their wings' vibrations in flight, and use them to transmit information and communicate among their species. The wings vibration sounds not only can reflect the species-specificity of fruit fly, but also can have advantage to fruit flies pests trapping and killing. Therefore, it has great biological importance and practical significance to research on fruit fly wings vibration sounds.
Scholars discovered courtship song among wings vibration sounds of fruit flies, which has very significant role in fruit flies capture and training. But the related research, which is about the classification of strains and gender through their wings vibration sounds, is opposite less, not comprehensive. So, in the article, wings vibration sounds of fruit flies are used as research object, and pattern recognition methods are tried to implement the classification of strains and gender of fruit flies. According to the experiment, we discover that the article's method is feasible and effective to strains identification and gender classification of fruit flies and provides a new method for fruit flies classification research.
The article's research mainly includes the following aspects:
(1) Summarizing research situations on fruit flies in home and abroad and problems existing in current research.
(2) Introducing the recognition principles and algorithms about Hidden Markov Model, and reviewing the current study status that Hidden Markov Model is used for animal sounds.
(3) Achieving the recognition for wings vibration sounds of three different strains of fruit flies. Firstly, it is introduced that the basic principle of Mel-frequency Cepstrum Coefficients method and the combination method of Gaussian Mixture Model and Hidden Markov Model. Secondly, according to the experimental data in the paper, the optimal parameters are selected for Mel-frequency Cepstrum Coefficients method, the effective features combination of wings vibration sounds of different strains of fruit flies, which is made of the energy feature and first-order variance characteristics, is extracted. Finally, models are established for wings vibration sounds of each strain fruit flies using the combination model with Gaussian Mixture Model and Hidden Markov Model, the wings vibration sound of unknown strain fruit fly is identified according to the max probability value of observation produced, and the recognition results are calculated. Experimental results show that the article's method is effective to the recognition of different strains fruit flies, and the results are good.
(4) Realizing the gender classification between female and male for wings vibration sounds of two strains of fruit flies. In the first, the theory of Auto-Regressive Model and Support Vector Machine are deeply studied. Then, combined with the experimental data, the appropriate parameters and algorithms of Auto-Regressive Model are selected, the orders are calculated by AIC criterion, the power spectrums of wings vibration sounds of female fruit flies and male fruit flies, which are the same strain, are estimated through Burg algorithm, the features combination, which can optimally reflect characteristics of wings vibration sounds of different gender fruit flies, includes the coefficients of Auto-Regressive Model and white noise sequence variance. Finally, the optimal parameters, which are suitable to the classifier of Support Vector Machine, are got by the ten fold cross-validation method, and using the heavy-tailed radial basis function as the kernel function of Support Vector Machine, the optimal classification interface is obtained through training sets, the testing set is classified, and the results are computed. Experimental results show that the article's method is feasible and effective for the classification between female and male in the same strain of fruit fly, and the results are good.
引文
[1]张鹏飞,毛玉蕊,杨绕华等.黑腹果蝇的一些生物学特性观察[J].北京农学院学报,2010,25(1)：16-19.
[2]刘蓓蓓.基于机器视觉的果蝇复眼病变图像识别系统的研究[D].湖南长沙：中南大学,2008.
[3]聂晓颖.果蝇鸣声特征提取及人工神经网络分类研究[D].陕西西安：陕西师范大学,2007.
[4]E. Tauber, D. F. Eberl. Acoustic Communication in Drosophila[J]. Behavioral Processes,2003,64:197-210.
[5]D. E. Cowling, B. Burnet. Courtship Song and Genetic Control of Their Acoustic Characteristics in Sibling Species of the Drosophila Melanogaster Subgroup[J]. Anim. Behav.,1981,29:924-935.
[6]陈暨耀,吴伟忠,蔡怀新等.用新的计量方法研究黑腹果蝇的求爱歌[J].动物学研究,1988,9(2)：133-139.
[7]庚镇诚,朱定良,孙耀来等.果蝇亚群中六个种的求爱歌的研究-对ipi作用的研究[J].遗传学报,1989,16(6)：448-454.
[8]J. Sivinski, C. O. Calkins, J. C. Webb. Comparisons of Acoustic Courtship Signals in Wild and Laboratory Reared Mediterranean Fruit Fly Certainties Capitata[J]. Florida Entomol,1989,72(1):212-214.
[9]袁越,王隽奇,钟忠等.时域-频域结合分析法-一种分析果蝇求爱歌的新方法[J].遗传学报,1992,19(6)：497-509.
[10]李成林,刘江伟,刘镇清等.果蝇求爱歌的声学分析[J].声学技术,1996,15(1)：29-30.
[11]邵红光,里敦,张咸宁等.果蝇nasuta亚群求爱歌的种间识别与进化遗传学研究[J].遗传学报,1997,24(4)：311-321.
[12]A. Mizrach, A. Hetzroni, M. Mazor, et al. Acoustic Trap For Female Mediterranean Fruit Flies[J]. American Society of Agricultural Engineers,2005, 48(5):2017-2022.
[13]J. C. Webb, C. O. Calkins, D. L. Chambers, et al. Acoustic Aspects of Behaviour of Mediterranean Fruit Fly, Ceratitis Capitata:Analysis and Identification of Coutship Sound[J]. Exp. Appl.,1983,33(1):1-8.
[14]J. C. Webb, J. L. Sharp, D. L. Chambers, et al. Acoustical Properties of the Flight Activities of the Caribbean Fruit Fly[J]. J. exp. Biol.,1976,64:761-772.
[15]C. P. Kyriacou, J. C. Hall. The Function of Courtship Song Rhythms in Drosophila[J]. Anim. Behav.,1982,30:794-801.
[16]C. P. Kyriacou, J. C. Hall. Learning and Memory Mutations Impair Acoustic Priming of Mating Behavior in Drosophila[J]. Nature,1984,308:62-65.
[17]C. P. Kyriacou, J. C. Hall. Interspecific Genetic Control of Courtship Song Production and Reception in Drosophila[J]. Science,1986,232:494-497.
[18]吴伟忠,陈暨耀,蔡怀新等.用微机技术测量黑腹果蝇求偶歌节律[J].复旦学报(自然科学版),1988,27(4)：402-406.
[19]聂晓颖,郭敏,何建平.人工神经网络对果蝇鸣声的分类识别[J].西北农林科技大学学报(自然科学版),2007,35(12)：201-204.
[20]L. R. Rabiner. A Tutorial on Hidden Markov Models and Selected Applications in Speech Recognition[J]. Proceedings of the IEEE,1989,77(2):257-286.
[21]L. R. Rabiner, B. H. Juang. An Introduction to Hidden Markov Models[J]. IEEE ASSP MAGAZINE,1986, January:4-16.
[22]B. A. Weisburn, S. G. Mitchell, C. W. Clark, et al. Isolating Biological Acoustic Transient Signals[C]. In Proceedings of the 1993 International Conference on Acoustics, Speech, and Signal Processing, Minneapolis, MN, USA,1993.
[23]J. A. Kogan, D. Margoliash. Automated Recognition of Bird Song Elements from Continuous Recordings Using Dynamic Time Warping and Hidden Markov Models:A Comparative Study[J]. Journal of the American Acoustical Society, 1998, vol.103.
[24]S. E. Anderson. Speech Recognition Meets Bird Song:a Comparison of Statistics-based and Template-based Techniques[J]. J. Acoust. Soc. Am.,1999,106: 2130.
[25]K. Adi, M. T. Johnson. Automatic Song-type Classification and Individual Identification of Ortolan Bunting (Emberiza Hortulana L) Vocalizations[C]. Fall 2004 Meetings of the Acoustical Society of America, San Diego,2004.
[26]M. T. Trawicki, M. T. Johnson. Automatic Song-Type Classification and Speaker Identification of Norwegian Ortolan Buting (Emberiza Hortulana) Vocalizations[C]. IEEE International Conference on Machine Learning in Signal Processing (MLSP), Mystic Connecticut, September 2005.
[27]P. J. Clemins. Automatic Classification of Animal Vocalizations[D]. Milwaukee, Wisconsin:Marquette University,2005.
[28]P. J. Clemins, M. T. Johnson, K. M. Leong, et al. Automatic Classification and Speaker Identification of African Elephant (Loxodonta Africana) vocalizations[J]. Journal of the Acoustical Society of America,2005, Vol.117(2):956-963.
[29]C. K. Adi. Hidden Markov Model Based Animal Acoustic Censusing:Learning From Speech Processing Technology[D]. Milwaukee, Wisconsin:Marquette University,2008.
[30]Y. Ren, M. T. Johnson, P. J. Clemins, et al. A Framework for Bioacoustic Vocalization Analysis Using Hidden Markov Models[J]. Algorithms,2009, vol.2 (3):1410-1428.
[31]J. C. Brown, P. Smaragdis. Automatic Classification of Vocalizations with Gaussian Mixture Models and Hidden Markov Models[J]. The Journal of the Acoustical Society of America,2008,123(5):3345-3345.
[32]卢鸣.HMM基本原理及其在聚类中的应用[D].江苏无锡：江南大学,2007.
[33]谢锦辉等.HMM及其在语音处理中的应用[M].湖北武汉：华中理工大学出版社,1995.
[34]张长胜.HMM在语音识别中的应用研究[D].吉林长春：吉林大学,2006.
[35]汤玲.基于HMM模型的语音识别系统研究[D].湖南长沙：国防科学技术大学,2005.
[36]P. Bansal, A. kant, and S. Kumar, el at. Improved Hybrid Model of HMM/GMM for Speech Recognition[J]. In Intelligent Information and Engineering Systems, 2008,69-74.
[37]X. D. Huang, A. Acero, H. W. Hon. Spoken Language Processing:a Guide to Theory, Algorithm, and System Development[M]. Upper Saddle River, New Jersey:Prentice Hall PTR,2001.
[38]吕霄云.基于MFCC和GMM的异常声音识别算法研究[D].四川成都：西南交通大学,2007.
[39]黄英来.基于动物声音的个体辨认技术的研究[D].哈尔滨：东北农业大学,2006.
[40]卢坚,陈毅松,孙正兴等.基于隐马尔科夫模型的音频自动分类[J].软件学报,2002,13(8)：1593-1597.
[41]陈程.基于HMM的语音识别系统研究[D].湖南长沙：中南大学,2008.
[42]I. Guler, M. K. Kiymik, M. Akin, et al. AR Spectral Analysis of EEG Signals by Using Maximum Likelihood Estimation[J]. Computers in Biology and Medicine, 2001,31:441-450.
[43]皇甫堪等.现代数字信号处理[M].北京：电子工业出版社,2003.
[44]边肇祺,张学工等.模式识别(第二版)[M].北京：清华大学出版社,2000.
[45]C. W. Hsu, C. C. Chang, C. J. Lin. A Practical Guide to Support Vector Classification[J/OL]. http://www.csie.ntu.edu.tw/～cjlin,2003.
[46]Chris. J. C. Burges. A Tutorial on Support Vector Machines for Pattern Recognition[J]. Data Mining and Knowledge Discovery,1998,2:121-167.
[47]P.H. Chen, C. J. Lin, B. Scholkopf. A Tutorial on v-Support Vector Machine[J/OL]. http://www.csie.ntu.edu.tw/cjlin/papers/nusvmtotorial.pdf,2003.
[48]O. Chapelle, P. Haffner, V. N. Vapnik. Support Vector Machines for Histogram-Based Image Classification[J]. IEEE TRANSACTIONS ON NEURAL NETWORKS,1999,10(5):1055-1064.
[49]E. D. ubeyli, D. Cvetkovic, G. Holland, et al. Analysis of Sleep EEG Activity during Hypopnoea Episodes by Least Squares Support Vector Machine Employing AR Coefficients[J]. Expert Systems with Applications,2010,37:4463-4467.
[50]王强.基于HHT和SVM的水电机组特征提取和状态识别[D].湖北武汉：华中科技大学,2008.
[51]王占强.基于SVM的模型选择和参数优化方案研究与实现[D].湖北武汉：华中科技大学,2008.
[2]刘蓓蓓.基于机器视觉的果蝇复眼病变图像识别系统的研究[D].湖南长沙：中南大学,2008.
[3]聂晓颖.果蝇鸣声特征提取及人工神经网络分类研究[D].陕西西安：陕西师范大学,2007.
[4]E. Tauber, D. F. Eberl. Acoustic Communication in Drosophila[J]. Behavioral Processes,2003,64:197-210.
[5]D. E. Cowling, B. Burnet. Courtship Song and Genetic Control of Their Acoustic Characteristics in Sibling Species of the Drosophila Melanogaster Subgroup[J]. Anim. Behav.,1981,29:924-935.
[6]陈暨耀,吴伟忠,蔡怀新等.用新的计量方法研究黑腹果蝇的求爱歌[J].动物学研究,1988,9(2)：133-139.
[7]庚镇诚,朱定良,孙耀来等.果蝇亚群中六个种的求爱歌的研究-对ipi作用的研究[J].遗传学报,1989,16(6)：448-454.
[8]J. Sivinski, C. O. Calkins, J. C. Webb. Comparisons of Acoustic Courtship Signals in Wild and Laboratory Reared Mediterranean Fruit Fly Certainties Capitata[J]. Florida Entomol,1989,72(1):212-214.
[9]袁越,王隽奇,钟忠等.时域-频域结合分析法-一种分析果蝇求爱歌的新方法[J].遗传学报,1992,19(6)：497-509.
[10]李成林,刘江伟,刘镇清等.果蝇求爱歌的声学分析[J].声学技术,1996,15(1)：29-30.
[11]邵红光,里敦,张咸宁等.果蝇nasuta亚群求爱歌的种间识别与进化遗传学研究[J].遗传学报,1997,24(4)：311-321.
[12]A. Mizrach, A. Hetzroni, M. Mazor, et al. Acoustic Trap For Female Mediterranean Fruit Flies[J]. American Society of Agricultural Engineers,2005, 48(5):2017-2022.
[13]J. C. Webb, C. O. Calkins, D. L. Chambers, et al. Acoustic Aspects of Behaviour of Mediterranean Fruit Fly, Ceratitis Capitata:Analysis and Identification of Coutship Sound[J]. Exp. Appl.,1983,33(1):1-8.
[14]J. C. Webb, J. L. Sharp, D. L. Chambers, et al. Acoustical Properties of the Flight Activities of the Caribbean Fruit Fly[J]. J. exp. Biol.,1976,64:761-772.
[15]C. P. Kyriacou, J. C. Hall. The Function of Courtship Song Rhythms in Drosophila[J]. Anim. Behav.,1982,30:794-801.
[16]C. P. Kyriacou, J. C. Hall. Learning and Memory Mutations Impair Acoustic Priming of Mating Behavior in Drosophila[J]. Nature,1984,308:62-65.
[17]C. P. Kyriacou, J. C. Hall. Interspecific Genetic Control of Courtship Song Production and Reception in Drosophila[J]. Science,1986,232:494-497.
[18]吴伟忠,陈暨耀,蔡怀新等.用微机技术测量黑腹果蝇求偶歌节律[J].复旦学报(自然科学版),1988,27(4)：402-406.
[19]聂晓颖,郭敏,何建平.人工神经网络对果蝇鸣声的分类识别[J].西北农林科技大学学报(自然科学版),2007,35(12)：201-204.
[20]L. R. Rabiner. A Tutorial on Hidden Markov Models and Selected Applications in Speech Recognition[J]. Proceedings of the IEEE,1989,77(2):257-286.
[21]L. R. Rabiner, B. H. Juang. An Introduction to Hidden Markov Models[J]. IEEE ASSP MAGAZINE,1986, January:4-16.
[22]B. A. Weisburn, S. G. Mitchell, C. W. Clark, et al. Isolating Biological Acoustic Transient Signals[C]. In Proceedings of the 1993 International Conference on Acoustics, Speech, and Signal Processing, Minneapolis, MN, USA,1993.
[23]J. A. Kogan, D. Margoliash. Automated Recognition of Bird Song Elements from Continuous Recordings Using Dynamic Time Warping and Hidden Markov Models:A Comparative Study[J]. Journal of the American Acoustical Society, 1998, vol.103.
[24]S. E. Anderson. Speech Recognition Meets Bird Song:a Comparison of Statistics-based and Template-based Techniques[J]. J. Acoust. Soc. Am.,1999,106: 2130.
[25]K. Adi, M. T. Johnson. Automatic Song-type Classification and Individual Identification of Ortolan Bunting (Emberiza Hortulana L) Vocalizations[C]. Fall 2004 Meetings of the Acoustical Society of America, San Diego,2004.
[26]M. T. Trawicki, M. T. Johnson. Automatic Song-Type Classification and Speaker Identification of Norwegian Ortolan Buting (Emberiza Hortulana) Vocalizations[C]. IEEE International Conference on Machine Learning in Signal Processing (MLSP), Mystic Connecticut, September 2005.
[27]P. J. Clemins. Automatic Classification of Animal Vocalizations[D]. Milwaukee, Wisconsin:Marquette University,2005.
[28]P. J. Clemins, M. T. Johnson, K. M. Leong, et al. Automatic Classification and Speaker Identification of African Elephant (Loxodonta Africana) vocalizations[J]. Journal of the Acoustical Society of America,2005, Vol.117(2):956-963.
[29]C. K. Adi. Hidden Markov Model Based Animal Acoustic Censusing:Learning From Speech Processing Technology[D]. Milwaukee, Wisconsin:Marquette University,2008.
[30]Y. Ren, M. T. Johnson, P. J. Clemins, et al. A Framework for Bioacoustic Vocalization Analysis Using Hidden Markov Models[J]. Algorithms,2009, vol.2 (3):1410-1428.
[31]J. C. Brown, P. Smaragdis. Automatic Classification of Vocalizations with Gaussian Mixture Models and Hidden Markov Models[J]. The Journal of the Acoustical Society of America,2008,123(5):3345-3345.
[32]卢鸣.HMM基本原理及其在聚类中的应用[D].江苏无锡：江南大学,2007.
[33]谢锦辉等.HMM及其在语音处理中的应用[M].湖北武汉：华中理工大学出版社,1995.
[34]张长胜.HMM在语音识别中的应用研究[D].吉林长春：吉林大学,2006.
[35]汤玲.基于HMM模型的语音识别系统研究[D].湖南长沙：国防科学技术大学,2005.
[36]P. Bansal, A. kant, and S. Kumar, el at. Improved Hybrid Model of HMM/GMM for Speech Recognition[J]. In Intelligent Information and Engineering Systems, 2008,69-74.
[37]X. D. Huang, A. Acero, H. W. Hon. Spoken Language Processing:a Guide to Theory, Algorithm, and System Development[M]. Upper Saddle River, New Jersey:Prentice Hall PTR,2001.
[38]吕霄云.基于MFCC和GMM的异常声音识别算法研究[D].四川成都：西南交通大学,2007.
[39]黄英来.基于动物声音的个体辨认技术的研究[D].哈尔滨：东北农业大学,2006.
[40]卢坚,陈毅松,孙正兴等.基于隐马尔科夫模型的音频自动分类[J].软件学报,2002,13(8)：1593-1597.
[41]陈程.基于HMM的语音识别系统研究[D].湖南长沙：中南大学,2008.
[42]I. Guler, M. K. Kiymik, M. Akin, et al. AR Spectral Analysis of EEG Signals by Using Maximum Likelihood Estimation[J]. Computers in Biology and Medicine, 2001,31:441-450.
[43]皇甫堪等.现代数字信号处理[M].北京：电子工业出版社,2003.
[44]边肇祺,张学工等.模式识别(第二版)[M].北京：清华大学出版社,2000.
[45]C. W. Hsu, C. C. Chang, C. J. Lin. A Practical Guide to Support Vector Classification[J/OL]. http://www.csie.ntu.edu.tw/～cjlin,2003.
[46]Chris. J. C. Burges. A Tutorial on Support Vector Machines for Pattern Recognition[J]. Data Mining and Knowledge Discovery,1998,2:121-167.
[47]P.H. Chen, C. J. Lin, B. Scholkopf. A Tutorial on v-Support Vector Machine[J/OL]. http://www.csie.ntu.edu.tw/cjlin/papers/nusvmtotorial.pdf,2003.
[48]O. Chapelle, P. Haffner, V. N. Vapnik. Support Vector Machines for Histogram-Based Image Classification[J]. IEEE TRANSACTIONS ON NEURAL NETWORKS,1999,10(5):1055-1064.
[49]E. D. ubeyli, D. Cvetkovic, G. Holland, et al. Analysis of Sleep EEG Activity during Hypopnoea Episodes by Least Squares Support Vector Machine Employing AR Coefficients[J]. Expert Systems with Applications,2010,37:4463-4467.
[50]王强.基于HHT和SVM的水电机组特征提取和状态识别[D].湖北武汉：华中科技大学,2008.
[51]王占强.基于SVM的模型选择和参数优化方案研究与实现[D].湖北武汉：华中科技大学,2008.