面向气象语音呼叫中心的语音合成软件设计与实现
|
摘要
呼叫中心的建立,已在越来越多的行业中崭露头角。目前168,400,800电话作为客服或咨询电话,被社会广泛的接受,呼叫中心也逐渐成为企业宣传、服务的重要途径之一,成为客户与企业沟通的重要纽带。国内的语音技术迅速发展,相关的语音产品也随之而出。语音产品能有长足的发展,数字语音技术的进步功不可没。
小波变换作为能随频率的变化自动调整分析窗大小的分析工具,自上世纪八十年代中期以来得到了迅猛的发展,并在诸如计算机视觉、计算机图形学、曲线设计、生物医学方面等众多的领域得到应用。通过小波理论,我们可以通过小波变换后得到的各个子波作适当的阈值处理,将小于阈值的小波系数置零,而保留大于阈值的小波系数,从而使含噪信号中的噪声得到有效的抑制。
本文围绕气象语音呼叫中心语音自动合成展开研究。简述了全国气象语音服务的背景,及本市的气象语音服务现状。结合小波经典理论,比较了传统的硬阈值算法和软阈值算法及新的小波阈值处理方法的优劣,在MatLab的模拟试验中对结论加之证实。并在小波去噪的基础上,在现有的硬件软件平台上,设计实现了适合日常化运作的语音自动合成软件。本文从新软件语音库的建立,分词法的选择,关键模块的功能,进行了详细的分析说明,并由此过程,总结工作并对未来的工作更深入的研究提出方向。
The establishment of call centers has been an increasing number of industries emerge. Currently, as 168,400,800 telephone customer service or advice phone, widely accepted by society, call centers have gradually become a one of important way to business advocacy and service.It has been a important way to link customer to corporate.China's speech technology has rapid developed,and related products are also appeared. Voice products have made great progress in help of digital voice and technological progress.
Wavelet transform as the frequency changes can be automatically adjusted with the analysis window size of the analytical tools, since the mid-eighties of the last century has been rapid development, such as computer vision, computer graphics, curve design, turbulence, biological medicine, and many other areas to be applied. Through the wavelet theory,we can get the various sub-waves to make appropriate threshold processing, we may process the subwaves by properly threshold algorithrm which set coefficients lower than the threshold to zero and keep the coefficients greater than the threshold. It will restrain effectively the noise.
Focusing on the research of voice weather service system, this article gives out a brief introduction of the background of national voice weather service as well as the locally current service situation. Combination of classical wavelet theory, compared to conventional hard threshold and soft threshold method and a new wavelet thresholding approach and the conclusions proved by the MatLab simulation test.Wavelet de-noising system based on the existing hardware and software platform, designed and implemented for the daily operation of automatic speech synthesis software.The article, from which the establishment of speech database, sub-lexical choice, the key modules, to carry out a detailed analysis, and from the process, review the work and put forward a propose directions in more in-depth study.
小波变换作为能随频率的变化自动调整分析窗大小的分析工具,自上世纪八十年代中期以来得到了迅猛的发展,并在诸如计算机视觉、计算机图形学、曲线设计、生物医学方面等众多的领域得到应用。通过小波理论,我们可以通过小波变换后得到的各个子波作适当的阈值处理,将小于阈值的小波系数置零,而保留大于阈值的小波系数,从而使含噪信号中的噪声得到有效的抑制。
本文围绕气象语音呼叫中心语音自动合成展开研究。简述了全国气象语音服务的背景,及本市的气象语音服务现状。结合小波经典理论,比较了传统的硬阈值算法和软阈值算法及新的小波阈值处理方法的优劣,在MatLab的模拟试验中对结论加之证实。并在小波去噪的基础上,在现有的硬件软件平台上,设计实现了适合日常化运作的语音自动合成软件。本文从新软件语音库的建立,分词法的选择,关键模块的功能,进行了详细的分析说明,并由此过程,总结工作并对未来的工作更深入的研究提出方向。
The establishment of call centers has been an increasing number of industries emerge. Currently, as 168,400,800 telephone customer service or advice phone, widely accepted by society, call centers have gradually become a one of important way to business advocacy and service.It has been a important way to link customer to corporate.China's speech technology has rapid developed,and related products are also appeared. Voice products have made great progress in help of digital voice and technological progress.
Wavelet transform as the frequency changes can be automatically adjusted with the analysis window size of the analytical tools, since the mid-eighties of the last century has been rapid development, such as computer vision, computer graphics, curve design, turbulence, biological medicine, and many other areas to be applied. Through the wavelet theory,we can get the various sub-waves to make appropriate threshold processing, we may process the subwaves by properly threshold algorithrm which set coefficients lower than the threshold to zero and keep the coefficients greater than the threshold. It will restrain effectively the noise.
Focusing on the research of voice weather service system, this article gives out a brief introduction of the background of national voice weather service as well as the locally current service situation. Combination of classical wavelet theory, compared to conventional hard threshold and soft threshold method and a new wavelet thresholding approach and the conclusions proved by the MatLab simulation test.Wavelet de-noising system based on the existing hardware and software platform, designed and implemented for the daily operation of automatic speech synthesis software.The article, from which the establishment of speech database, sub-lexical choice, the key modules, to carry out a detailed analysis, and from the process, review the work and put forward a propose directions in more in-depth study.
引文
[1]张家騄,齐士钤,俞舸.汉语语音合成系统评价方法.声学学报,1998,(01).
[2]Stephane Mallat,杨力华等译.信号处理的小波导引.北京：机械工业出版社,2003,9
[3]陈峰,成新民.基于小波变换的信号去噪技术及实现.现代电子技术.2005,3,11-13
[4]周密,李尊尊,耿国华.基于小波阈值的图像去噪方法研究.计算机技术与发展,2008,18(5)：22-24
[5]张翠芳.小波阈值降噪效果影响因素的研究.西安邮电学院学报.2008,13(5)：13-15
[6]韩震宇,申旭娟,石章林.信号的多分辨率分析及其在消噪中的应用.四川联合大学学报,1999,3：30-31
[7]蔡云泽,文成林.一种新的基于多尺度分析的滤波算法[J].工程数学学报,2000,17(4)：50-54
[8]文鸿雁,张正禄.非线性小波变换阈值法去噪改进.测绘通报,2006,3：18-21
[9]仇智华,艾德才.基于子波分析的信号噪声分离方法研究.天津大学学报,2000,33：45-48
[10]葛哲学,陈仲生.Matlab时频分析技术及其应用.北京：人民邮电出版社,2006,1
[11]王一世.数字信号处理.北京：理工大学出版社,1997,pp.180-188
[12]赵焕东,李志能,沈会良.小波变换在语音信号处理中的运用.浙江大学学报工学版,2001,36：13-14
[13]郑海波,李志远.基于小波包变换的一种降噪算法.合肥工业大学学报(自然版),2001,24(6)：459-462
[14]沈亚强,金洪霞,刘旭.基于子波变换的语音去噪方法.信号处理,2000,16(3)：221-226
[15]李高峰,尹俊勋,徐艳.一种改进的子波域语音去噪方法.电声技术.2001,1(187)：9-11
[16]沈亚强,金洪震.一种基于子波变换的语音增强方法.科技通报,2000,16(3)：206-211
[17]胡国胜,陈一天,任震.Fourier变换和小波变换的一般形式.计算机工程与应用,2001,38：15-1
[18]马丽萍,余华端.含噪声混沌信号的小波去噪方法的研究.信号处理,2001,18：11-12
[19]Jansen M, Molfoit M, Bultheel A. Generalized Cross Validation for wavelet thresholding[J]. Signal Processing,1997,56(1):33-34
[20]Chang S G, Bin Yu, Vetlerli M. Adaptive wavelet thresholding for image denoising and compression[J], IEEE Trans Image Processing,2000, 9(9):1532-1546
[21]Bruce A G, Gao H Y. Waveshrink with firm shrinkage[M]. Statist. Sinica,1997, 7(4):855-874
[22]Divid L, Donoho. Denosing by soft-threshoulding. IEEE trans on IT,1995, 3(40):613-627
[23]Sha Piro J. Embedded image coding using zerotrees of wavelet coefficients. IEEE Trans. on Signal Proeessing.1993,41:3445-3462
[24]Soon Ing Yann, Koh Soo Ngee, Yeo Chai Kiat. Wavelet for Speech Denoising. Proceedings of IEEE Region 10 Annual International Conference,1997, 2:479-482
[25]Kain A, Macon M. Spectral Voice Conversion For Tex-to-Speech Synthesis. Proceedings of ICASSP, Seattle, Washington, May 1998:285-288
[26]Kain A, Macon M. Text-t-Speech voice adaptation from sparse training data. Proceedings of the ICSIP 1998, Sydney, Australia, December 1998,7:2847-2850
[27]Rentzos D, Vaseghi S, Yan Q, Ho C H. Voice conversion through transformation of spectral and intonation features. ICASSP,2004,I:21-24
[28]Rentzos D, Vaseghi S, Yan Q, Turajlic E. Probabiliy models of formant Pararmeter for voice conversion. EUROSPEECH 2003,2405-2408
[29]Stylianou Y. Applying the Harmonic Plus Noise Model in Concatenative Speech Synthesis. IEEE Transactionson Speech and Audio Processing, January 2001, 9(1):21-29
[30]Verhelst W, Mertens J. Voice Conversion Using Partions of Spectral Feature Space. ICCASSP, Atlanta USA, may 1996,1:365-368
[31]Wegmann S, McAllaster D, Orloff J, Peskin B. Speaker normalization conversational telephone speech. ICASSP, Atlanta, USA,1996
[32]Haitao Fang. Deshuang Huang.Noise reduction in lidar signal based on discrete wavelet transform. Optics Communications,2004,(233):67-76
[33]Taichiu Hsung, Daniel Pakkong Lun, K C Ho. Optimizing the multiwavelet shrinkage de-noising. IEEE Trans. on signal processing.2005,53(1):240-251
[34]T R Downie, B W Silverman. The Discrete Multiple Wavelet Transform and Thresholding methods. IEEE Trans on signal processing,1998,46(9):2558-2561
[35]Zhen Xianlin, Guo xiangsong, Wen xue. Two improved methods on wavelet image denoising. Proceedings of the Second International Conference on Machine Learding and Cybernetics,2003, (11):2979-2
[2]Stephane Mallat,杨力华等译.信号处理的小波导引.北京：机械工业出版社,2003,9
[3]陈峰,成新民.基于小波变换的信号去噪技术及实现.现代电子技术.2005,3,11-13
[4]周密,李尊尊,耿国华.基于小波阈值的图像去噪方法研究.计算机技术与发展,2008,18(5)：22-24
[5]张翠芳.小波阈值降噪效果影响因素的研究.西安邮电学院学报.2008,13(5)：13-15
[6]韩震宇,申旭娟,石章林.信号的多分辨率分析及其在消噪中的应用.四川联合大学学报,1999,3：30-31
[7]蔡云泽,文成林.一种新的基于多尺度分析的滤波算法[J].工程数学学报,2000,17(4)：50-54
[8]文鸿雁,张正禄.非线性小波变换阈值法去噪改进.测绘通报,2006,3：18-21
[9]仇智华,艾德才.基于子波分析的信号噪声分离方法研究.天津大学学报,2000,33：45-48
[10]葛哲学,陈仲生.Matlab时频分析技术及其应用.北京：人民邮电出版社,2006,1
[11]王一世.数字信号处理.北京：理工大学出版社,1997,pp.180-188
[12]赵焕东,李志能,沈会良.小波变换在语音信号处理中的运用.浙江大学学报工学版,2001,36：13-14
[13]郑海波,李志远.基于小波包变换的一种降噪算法.合肥工业大学学报(自然版),2001,24(6)：459-462
[14]沈亚强,金洪霞,刘旭.基于子波变换的语音去噪方法.信号处理,2000,16(3)：221-226
[15]李高峰,尹俊勋,徐艳.一种改进的子波域语音去噪方法.电声技术.2001,1(187)：9-11
[16]沈亚强,金洪震.一种基于子波变换的语音增强方法.科技通报,2000,16(3)：206-211
[17]胡国胜,陈一天,任震.Fourier变换和小波变换的一般形式.计算机工程与应用,2001,38：15-1
[18]马丽萍,余华端.含噪声混沌信号的小波去噪方法的研究.信号处理,2001,18：11-12
[19]Jansen M, Molfoit M, Bultheel A. Generalized Cross Validation for wavelet thresholding[J]. Signal Processing,1997,56(1):33-34
[20]Chang S G, Bin Yu, Vetlerli M. Adaptive wavelet thresholding for image denoising and compression[J], IEEE Trans Image Processing,2000, 9(9):1532-1546
[21]Bruce A G, Gao H Y. Waveshrink with firm shrinkage[M]. Statist. Sinica,1997, 7(4):855-874
[22]Divid L, Donoho. Denosing by soft-threshoulding. IEEE trans on IT,1995, 3(40):613-627
[23]Sha Piro J. Embedded image coding using zerotrees of wavelet coefficients. IEEE Trans. on Signal Proeessing.1993,41:3445-3462
[24]Soon Ing Yann, Koh Soo Ngee, Yeo Chai Kiat. Wavelet for Speech Denoising. Proceedings of IEEE Region 10 Annual International Conference,1997, 2:479-482
[25]Kain A, Macon M. Spectral Voice Conversion For Tex-to-Speech Synthesis. Proceedings of ICASSP, Seattle, Washington, May 1998:285-288
[26]Kain A, Macon M. Text-t-Speech voice adaptation from sparse training data. Proceedings of the ICSIP 1998, Sydney, Australia, December 1998,7:2847-2850
[27]Rentzos D, Vaseghi S, Yan Q, Ho C H. Voice conversion through transformation of spectral and intonation features. ICASSP,2004,I:21-24
[28]Rentzos D, Vaseghi S, Yan Q, Turajlic E. Probabiliy models of formant Pararmeter for voice conversion. EUROSPEECH 2003,2405-2408
[29]Stylianou Y. Applying the Harmonic Plus Noise Model in Concatenative Speech Synthesis. IEEE Transactionson Speech and Audio Processing, January 2001, 9(1):21-29
[30]Verhelst W, Mertens J. Voice Conversion Using Partions of Spectral Feature Space. ICCASSP, Atlanta USA, may 1996,1:365-368
[31]Wegmann S, McAllaster D, Orloff J, Peskin B. Speaker normalization conversational telephone speech. ICASSP, Atlanta, USA,1996
[32]Haitao Fang. Deshuang Huang.Noise reduction in lidar signal based on discrete wavelet transform. Optics Communications,2004,(233):67-76
[33]Taichiu Hsung, Daniel Pakkong Lun, K C Ho. Optimizing the multiwavelet shrinkage de-noising. IEEE Trans. on signal processing.2005,53(1):240-251
[34]T R Downie, B W Silverman. The Discrete Multiple Wavelet Transform and Thresholding methods. IEEE Trans on signal processing,1998,46(9):2558-2561
[35]Zhen Xianlin, Guo xiangsong, Wen xue. Two improved methods on wavelet image denoising. Proceedings of the Second International Conference on Machine Learding and Cybernetics,2003, (11):2979-2