基于奇异值分解和小波变换的音频数字水印研究
|
摘要
计算机和网络的广泛应用,大大地方便了人们获取信息和交流信息,同时其版权保护也变得越来越重要。而数字水印技术作为一种有效的版权保护手段越来越受到人们的青睐。近年来,数字水印研究大多是基于静态图像的水印嵌入研究,对音频水印的研究较少。这是因为人的视觉和听觉的特性差别比较大,在音频信号中嵌入水印相对较难。
本论文以二值图像作为水印信息,以音频信号作为载体数据,以奇异值分解和离散小波变换理论为基础,提出了两种鲁棒性音频数字水印方法。
一是对原始音频信号进行离散小波变换,将提取的低频近似分量和二值水印图像都进行奇异值分解,分别获得奇异值对角阵向量。然后借助水印图像的奇异值对角阵向量修改低频近似分量的奇异值对角阵向量,实现水印的嵌入。利用流行音乐、古典音乐和语音三种音频样本分别进行鲁棒性测试,采用归一化相关系数和信噪比对水印和音频信号进行定量评价。仿真实验表明,三种音频样本对多种攻击具有较好的鲁棒性。但是此算法在水印提取时仍然需要原始音频信息,使其在实际应用中受到限制。
二是对前一种不是盲水印方法而进行的改进,是将原始音频数据经过离散小波变换后,提取低频分量分块进行奇异值分解,根据水印信息的值和设定的步长,来量化选取的奇异值,实现水印的嵌入。在水印嵌入前对水印图像进行Arnold置乱处理,提高了水印的安全性。采用分段奇异值分解提高了水印嵌入速度,节省了不少的运行时间。仿真实验中通过与基于SVD水印嵌入方法的归一化相关系数和信噪比的比较,可以看出此水印方法除了在进行去噪攻击中,水印隐藏效果稍差一些以外,对其他多种攻击都具有较好的鲁棒性。特别是此方法利用了量化处理,使得水印在检测与提取过程中不再需要原始载体数据,这样更加利于实际中的应用。
总的来说,两种水印方法借助奇异值的稳定性,抑制了噪声对音频数据的影响。利用离散小波变换提高了水印的不可感知性,都是可行的数字音频水印方法。
For the widespread application of computer and internet, it is very convenient for us to obtain the information and communicate with each other. At the same time, it is much more important to protect the copyright of digital data. As an effectual means for protecting the copyright, the digital watermarking had been researched widely. In recent years, the research of digital watermarking was mostly based on the static images imbedded. Due to the big difference of vision and audition of human being, it is difficult to imbed digital watermarking into the audio signal.
In this paper, two kinds of robustness audio digital watermarking method were proposed, with the two-dimension binary image as watermark information, the audio data as carrier data, on the basis of Singular Value Decomposition and Discrete Wavelet Transform.
(1) The audio signal was decomposed by using appropriate wavelet basis. Low frequency coefficients were selected and SVD of the divided signals was made. At the same time, the watermark was constructed by Singular Value Decomposition , and obtained a singular value vector. Then with the aid of a watermark image singular value diagonal vector, watermark embedded process was realized. Three kind of audio samples were carried on the robust test, such as pop music, classical music and speech. It was be maked the quantitative evaluation with the normalized correlation coefficient and the signal-to-noise ratio. Experimental results showed that method in this chapter was robust to many attacks, such as resampling, low pass filtering, noise addition ,cut and so on. This algorithm shortcoming was that when the watermark was extracted, it needed the primitive watermark information. So it would be restricted in the practical application.
(2) The previous kind was not improved which the blind watermark algorithm had be carried on. First, before watermark embedded, the watermark image was carried on Arnold transformation in order to realize covert effect. Afterward, the original audio signal was decomposed using appropriate wavelet basis. Low frequency coefficients were selected to divide into section and SVD of the divided signals was made. The singular values were chosen and embedded into the watermark image by quantization method. The using of the scrambling encryption, provided the security of the watermark. The using of the partition singular value decomposition, enhanced the watermark inserting speed and saved many running time. Normalized correlation coefficient and signal-to-noise ratio were compared with based on SVD watermark embedded method, The results of experimentation showed the watermark hideaway effect was slightly bad except the denoise attack, others had the good robustness. In particular, the watermark could be extracted without the original digital audio signal, it would be in favor of the actual application.
In conclusion, it is mostly purpose that the singular value stability in the Singular Value Decomposition can control noise to audio data influence. The using of the Wavelet transform improves the imperceptible of the watermark.
本论文以二值图像作为水印信息,以音频信号作为载体数据,以奇异值分解和离散小波变换理论为基础,提出了两种鲁棒性音频数字水印方法。
一是对原始音频信号进行离散小波变换,将提取的低频近似分量和二值水印图像都进行奇异值分解,分别获得奇异值对角阵向量。然后借助水印图像的奇异值对角阵向量修改低频近似分量的奇异值对角阵向量,实现水印的嵌入。利用流行音乐、古典音乐和语音三种音频样本分别进行鲁棒性测试,采用归一化相关系数和信噪比对水印和音频信号进行定量评价。仿真实验表明,三种音频样本对多种攻击具有较好的鲁棒性。但是此算法在水印提取时仍然需要原始音频信息,使其在实际应用中受到限制。
二是对前一种不是盲水印方法而进行的改进,是将原始音频数据经过离散小波变换后,提取低频分量分块进行奇异值分解,根据水印信息的值和设定的步长,来量化选取的奇异值,实现水印的嵌入。在水印嵌入前对水印图像进行Arnold置乱处理,提高了水印的安全性。采用分段奇异值分解提高了水印嵌入速度,节省了不少的运行时间。仿真实验中通过与基于SVD水印嵌入方法的归一化相关系数和信噪比的比较,可以看出此水印方法除了在进行去噪攻击中,水印隐藏效果稍差一些以外,对其他多种攻击都具有较好的鲁棒性。特别是此方法利用了量化处理,使得水印在检测与提取过程中不再需要原始载体数据,这样更加利于实际中的应用。
总的来说,两种水印方法借助奇异值的稳定性,抑制了噪声对音频数据的影响。利用离散小波变换提高了水印的不可感知性,都是可行的数字音频水印方法。
For the widespread application of computer and internet, it is very convenient for us to obtain the information and communicate with each other. At the same time, it is much more important to protect the copyright of digital data. As an effectual means for protecting the copyright, the digital watermarking had been researched widely. In recent years, the research of digital watermarking was mostly based on the static images imbedded. Due to the big difference of vision and audition of human being, it is difficult to imbed digital watermarking into the audio signal.
In this paper, two kinds of robustness audio digital watermarking method were proposed, with the two-dimension binary image as watermark information, the audio data as carrier data, on the basis of Singular Value Decomposition and Discrete Wavelet Transform.
(1) The audio signal was decomposed by using appropriate wavelet basis. Low frequency coefficients were selected and SVD of the divided signals was made. At the same time, the watermark was constructed by Singular Value Decomposition , and obtained a singular value vector. Then with the aid of a watermark image singular value diagonal vector, watermark embedded process was realized. Three kind of audio samples were carried on the robust test, such as pop music, classical music and speech. It was be maked the quantitative evaluation with the normalized correlation coefficient and the signal-to-noise ratio. Experimental results showed that method in this chapter was robust to many attacks, such as resampling, low pass filtering, noise addition ,cut and so on. This algorithm shortcoming was that when the watermark was extracted, it needed the primitive watermark information. So it would be restricted in the practical application.
(2) The previous kind was not improved which the blind watermark algorithm had be carried on. First, before watermark embedded, the watermark image was carried on Arnold transformation in order to realize covert effect. Afterward, the original audio signal was decomposed using appropriate wavelet basis. Low frequency coefficients were selected to divide into section and SVD of the divided signals was made. The singular values were chosen and embedded into the watermark image by quantization method. The using of the scrambling encryption, provided the security of the watermark. The using of the partition singular value decomposition, enhanced the watermark inserting speed and saved many running time. Normalized correlation coefficient and signal-to-noise ratio were compared with based on SVD watermark embedded method, The results of experimentation showed the watermark hideaway effect was slightly bad except the denoise attack, others had the good robustness. In particular, the watermark could be extracted without the original digital audio signal, it would be in favor of the actual application.
In conclusion, it is mostly purpose that the singular value stability in the Singular Value Decomposition can control noise to audio data influence. The using of the Wavelet transform improves the imperceptible of the watermark.
引文
[1]Ahemd H,Tewfik.Algorithms for Digital Watermarking.A Thesis Submitted to the Faculty of the University of Minnesota,2003:1.
[2]Hamza Ozer,Bulent Sankur,Nasir Memon.An SVD-based audio watermarking technique[J].Proceedings of the 7th workshop on Multimedia and security,New York,NY,USA,2005,August 01-02,51-56.
[3]FuYu,Wang Bao-bao,Li Chun-ru,Quan Ning-qiang.A Novel Algorithm for Robust Audio Watermarking in Wavelet Domain[J].Joumal of Electronic Science and Technology of China,2004,Jun,Vo12(2):70-72,78.
[4]Podilchuk.C.I.(Bell Laboratories);Delp.E.J.Digital watermarking:Algorithm and application[J].IEEE Signal Processing Magazine,2001,July,v 18(4):33-46.
[5]王炳锡,陈琦,邓峰森.数宇水印技术[M].陕西:西安电子科技大学出版社,2003.
[6]Cox I J,Miller M.L.The first 50 years of electronic watermarking[J].EURASIPJ of Applied Signal Processing,2002,2:126-132.
[7]周新法.基于小波变换的同步音频数字水印技术的研究:[硕士学位论文].江苏大学,2005年.
[8]谈华斌.数字音频水印算法的研究:[硕士学位论文].吉林大学,2004年.
[9]I.J.Cox etal.Secure spectrum watermarking for multimedia[J].IEEE Trans.on Image Processing,1997,Vol.6(12):1673-1678.
[10]林福宗.多媒体技术基础[M].北京:清华大学出版社,2000年.
[11]王翀.数字音频水印技术的研究:[硕士学位论文].大连理工大学,2005年.
[12]Bender W,Gruhl D,Morimoto N and Lu A.Techniques for data hiding.IBM Systems Joumal,1996,35(3&4):313-336.
[13]Gruhl D,lu A,Bender W.Echo hiding.Information hiding:first international workshop.1996,Cambridge,Uk,1174:295-315.
[14]I.B.Ozer,M.Rarnkumar.A.N.Akansu.FFT based signaling for multimedia steganography[A].Proc.IEEE Int.Conf.On Acoustics,Speec.and Signal Processing[C].Istanbul,Turkey,June 2000.1979-1982.
[15]A.Piva.M.bami.F.Bartonlini.V.Cappellini.DCT based watermark recovering without resorting to the uncorrupted original image [A].Proc.IEEE Int.Conf.On Acoustics,Speech,and Signal Processing[C].Istanbul,Turkey,June 1997.520-523.
[16]陈琦,王炳锡.一种基于DCT变换的语音数字水印算法研究[J].信号处理,2002,17(3):238-241.
[17]于晓敏.基于离散小波变换的音频数字水印算法研究[J].齐齐哈尔大学报,2007,23(1):60-64.
[18]刘瑞祯,谭铁牛.基于奇异值分解的数字图像水印方法[J].电子学报,2001:29(2),168-171.
[19]王潇,柏森,赵波.基于奇异值分解与融合的音频盲认证算法[J].中山大学学报,2004,43(增刊2):188-191.
[20]李海峰,王树勋,温泉,宋巍巍.基于SVD和ICA的鲁棒水印算法[J].中山大学学报(自然科学版),2004,43(增2):53-57. 算法[J].微计算机信息,2006,22(8-3):250-253.
[22]徐达文,王让定.基于卷积码的盲音频水印算法研究[J].计算机应用,2006,26(7):1649-1651.
[23]马田,张新鹏,王朔中.数字音频信号中的频域扰动调制水印嵌入[J].信号处理,2002,18(3):2002-2007.
[24]N.D.Jayant,J.D.Johnston.and R.J.Safranek.Signal compression based on models of human perception.Proc.IEEE,Oct,1993,vol.81,1385-1422.
[25]王剑,林福宗.基于人工神经网络的数字音频水印算法[J].小型微型计算机系统,2004,25(11):2006-2010.
[26]Wang Chong,Ma Xiaohong,Cong Xiaogping etal.An audio watermarking scheme with neural network.The International Symposium on Neural Networks(ISNN2005),Chongqing,China,2005:795-800.
[27]Martin K,Febien A,Petitcolas P.A Fair Benchmark for Image Watermarking Systems.Proceeding of Electronic Imageing.Washington:SPIE Press,1999.226-239.
[28]王向阳,杨红颖,赵红.一种新的自适应量化数字音频水印算法[J].声学技术,2004,3:117-120.
[29]王泳,黄继武,Yun Q.shi.快速重同步的有意义音频水印盲检测算法[J].计算机研究与发展,2003,2:215-230.
[30]王向阳,杨红颖.基于离散余弦变换的自适应数字音频水印技术研究[J].小型微型计算机系统,2004,10:1825-1827.
[31]张开文,张新鹏,王朔中.图像及音频信号中隐蔽嵌入信息存在性的统计检验[J].电子与信息学报,2003,7:872-877.
[32]杨洋,陈小平.一种变换域的音频多水印算法探索[J].微电子学与计 算机,2004,6:63-66.
[33]梁华庆,赵丽丽,钮心忻,杨义先.一种基于心理声学模型的小波包域音频数字水印算法[J].石油大学学报(自然科学版),2003,6:112-115.
[34]王让定,徐达文,陈金儿.基于频率掩蔽效应的自适应音频数字水印技术[J].计算机工程与应用,2004,15:131-33.
[35]颜菲,季兵,章德.抑制原始信号影响的数字音频水印算法[J].计算机应用,2003,12:111-113.
[36]丛湘平.鲁棒数字音频水印算法的研究与实现:[硕士学位论文].大连理工大学,2005年.
[37]钮心忻.信息隐藏与数字水印[M].北京:北京邮电大学出版社,2004:155-156
[38]Christine I.Podilchuk and Edward J.Delp,”Digital watermarking algorithm and application”,IEEE Signal Processing Magazine,July 2001,1053-5888.
[39]魏明果.实用小波分析[M].北京:北京理工大学出版社,2005:16,25.
[40]胡昌华,张军波,夏军,张伟.基于MATLAB的系统分析与设计--小波分析[M].西安:西安电子科技大学出版社,1999.
[41]彭玉华小波变换与工程应用[M].北京:科学出版社,2000:38-39.
[42]冉启文.小波变换与分数傅立叶变换理论及应用[M].哈尔滨:哈尔滨工业大学出版社,2001.21-22.
[43]陈景良,陈向晖.特殊矩阵[M].北京:清华大学出版社,2001,328-333.
[44]丁纬,闫伟奇,齐东旭.基于Arnold变换的数字图像置乱技术[J].计算机辅助设计与图形学学报,2001,13(4):338-341.
[2]Hamza Ozer,Bulent Sankur,Nasir Memon.An SVD-based audio watermarking technique[J].Proceedings of the 7th workshop on Multimedia and security,New York,NY,USA,2005,August 01-02,51-56.
[3]FuYu,Wang Bao-bao,Li Chun-ru,Quan Ning-qiang.A Novel Algorithm for Robust Audio Watermarking in Wavelet Domain[J].Joumal of Electronic Science and Technology of China,2004,Jun,Vo12(2):70-72,78.
[4]Podilchuk.C.I.(Bell Laboratories);Delp.E.J.Digital watermarking:Algorithm and application[J].IEEE Signal Processing Magazine,2001,July,v 18(4):33-46.
[5]王炳锡,陈琦,邓峰森.数宇水印技术[M].陕西:西安电子科技大学出版社,2003.
[6]Cox I J,Miller M.L.The first 50 years of electronic watermarking[J].EURASIPJ of Applied Signal Processing,2002,2:126-132.
[7]周新法.基于小波变换的同步音频数字水印技术的研究:[硕士学位论文].江苏大学,2005年.
[8]谈华斌.数字音频水印算法的研究:[硕士学位论文].吉林大学,2004年.
[9]I.J.Cox etal.Secure spectrum watermarking for multimedia[J].IEEE Trans.on Image Processing,1997,Vol.6(12):1673-1678.
[10]林福宗.多媒体技术基础[M].北京:清华大学出版社,2000年.
[11]王翀.数字音频水印技术的研究:[硕士学位论文].大连理工大学,2005年.
[12]Bender W,Gruhl D,Morimoto N and Lu A.Techniques for data hiding.IBM Systems Joumal,1996,35(3&4):313-336.
[13]Gruhl D,lu A,Bender W.Echo hiding.Information hiding:first international workshop.1996,Cambridge,Uk,1174:295-315.
[14]I.B.Ozer,M.Rarnkumar.A.N.Akansu.FFT based signaling for multimedia steganography[A].Proc.IEEE Int.Conf.On Acoustics,Speec.and Signal Processing[C].Istanbul,Turkey,June 2000.1979-1982.
[15]A.Piva.M.bami.F.Bartonlini.V.Cappellini.DCT based watermark recovering without resorting to the uncorrupted original image [A].Proc.IEEE Int.Conf.On Acoustics,Speech,and Signal Processing[C].Istanbul,Turkey,June 1997.520-523.
[16]陈琦,王炳锡.一种基于DCT变换的语音数字水印算法研究[J].信号处理,2002,17(3):238-241.
[17]于晓敏.基于离散小波变换的音频数字水印算法研究[J].齐齐哈尔大学报,2007,23(1):60-64.
[18]刘瑞祯,谭铁牛.基于奇异值分解的数字图像水印方法[J].电子学报,2001:29(2),168-171.
[19]王潇,柏森,赵波.基于奇异值分解与融合的音频盲认证算法[J].中山大学学报,2004,43(增刊2):188-191.
[20]李海峰,王树勋,温泉,宋巍巍.基于SVD和ICA的鲁棒水印算法[J].中山大学学报(自然科学版),2004,43(增2):53-57. 算法[J].微计算机信息,2006,22(8-3):250-253.
[22]徐达文,王让定.基于卷积码的盲音频水印算法研究[J].计算机应用,2006,26(7):1649-1651.
[23]马田,张新鹏,王朔中.数字音频信号中的频域扰动调制水印嵌入[J].信号处理,2002,18(3):2002-2007.
[24]N.D.Jayant,J.D.Johnston.and R.J.Safranek.Signal compression based on models of human perception.Proc.IEEE,Oct,1993,vol.81,1385-1422.
[25]王剑,林福宗.基于人工神经网络的数字音频水印算法[J].小型微型计算机系统,2004,25(11):2006-2010.
[26]Wang Chong,Ma Xiaohong,Cong Xiaogping etal.An audio watermarking scheme with neural network.The International Symposium on Neural Networks(ISNN2005),Chongqing,China,2005:795-800.
[27]Martin K,Febien A,Petitcolas P.A Fair Benchmark for Image Watermarking Systems.Proceeding of Electronic Imageing.Washington:SPIE Press,1999.226-239.
[28]王向阳,杨红颖,赵红.一种新的自适应量化数字音频水印算法[J].声学技术,2004,3:117-120.
[29]王泳,黄继武,Yun Q.shi.快速重同步的有意义音频水印盲检测算法[J].计算机研究与发展,2003,2:215-230.
[30]王向阳,杨红颖.基于离散余弦变换的自适应数字音频水印技术研究[J].小型微型计算机系统,2004,10:1825-1827.
[31]张开文,张新鹏,王朔中.图像及音频信号中隐蔽嵌入信息存在性的统计检验[J].电子与信息学报,2003,7:872-877.
[32]杨洋,陈小平.一种变换域的音频多水印算法探索[J].微电子学与计 算机,2004,6:63-66.
[33]梁华庆,赵丽丽,钮心忻,杨义先.一种基于心理声学模型的小波包域音频数字水印算法[J].石油大学学报(自然科学版),2003,6:112-115.
[34]王让定,徐达文,陈金儿.基于频率掩蔽效应的自适应音频数字水印技术[J].计算机工程与应用,2004,15:131-33.
[35]颜菲,季兵,章德.抑制原始信号影响的数字音频水印算法[J].计算机应用,2003,12:111-113.
[36]丛湘平.鲁棒数字音频水印算法的研究与实现:[硕士学位论文].大连理工大学,2005年.
[37]钮心忻.信息隐藏与数字水印[M].北京:北京邮电大学出版社,2004:155-156
[38]Christine I.Podilchuk and Edward J.Delp,”Digital watermarking algorithm and application”,IEEE Signal Processing Magazine,July 2001,1053-5888.
[39]魏明果.实用小波分析[M].北京:北京理工大学出版社,2005:16,25.
[40]胡昌华,张军波,夏军,张伟.基于MATLAB的系统分析与设计--小波分析[M].西安:西安电子科技大学出版社,1999.
[41]彭玉华小波变换与工程应用[M].北京:科学出版社,2000:38-39.
[42]冉启文.小波变换与分数傅立叶变换理论及应用[M].哈尔滨:哈尔滨工业大学出版社,2001.21-22.
[43]陈景良,陈向晖.特殊矩阵[M].北京:清华大学出版社,2001,328-333.
[44]丁纬,闫伟奇,齐东旭.基于Arnold变换的数字图像置乱技术[J].计算机辅助设计与图形学学报,2001,13(4):338-341.