½ 33 ½ 3 Þ Vol. 33, No. 3 7 3 ACTA AUTOMATICA SINICA March, 7 è ¹ MP3 ß å 1, Ä 1 1 ý  Åè ó ó ß Ì ß ñ1) Ä Ǒ ² ÂÔÅ þíò) û Ð (Discrete wavelet transform, DWT) Ð ßÙ (Discrete cosine transform, DCT) Í Í Å Âò3) ¹ ó ùþđ ÂÔÅ Đ ò4) Ý Đ Ǒ ÝÅ Ù å Ä Â ß Đ Þ Í á Ð þ Ç» Ì ß Ù Ǒ MP3 þ Ö ¹ Í (ðô ³ ) ³ MP3 ÞĐ TP391 A Robust Adaptive Digital Audio Watermarking Scheme Against MP3 Compression WANG Xiang-Yang 1, YANG Hong-Ying 1 ZHAO Hong 1 Abstract In this paper, a robust adaptive digital audio watermarking scheme in the hybrid domain is presented. The features of the proposed algorithm are as follows: 1)More robust synchronization codes and a new embedded strategy are adopted to resist the attack effectively. )The multi-resolution characteristics of discrete wavelet transform(dwt) and the energy-compression characteristics of discrete cosine transform(dct) are combined to improve the transparency of digital watermark. 3)The adaptive quantization steps are identified according to human auditory masking. 4)The algorithm can extract the watermark without the help of the original digital audio signal. Experimental results show that the proposed watermarking algorithm is imperceptible and robust against various signal processings such as re-sampling, re-quantifying, noising adding, low pass filtering and cropping. Especially, our audio watermarking scheme is remarkably effective in resisting MP3 compression. Key words Digital watermarking, digital audio, MP3 compression, quantization step 1 Ý ß Ǒ þú 5-9- þ ÍÚ 6-6-3 Received September, 5; in revised form June 3, 6 Û (31) ü ² ÒÁ û ( ) Ñè (53), ù (6J3JH), Ç ² Ò û ( ) Ñèáà (KJS6), ä Ò ä ² û (À ) Ñè (ZK514) á Support by the Natural Science Foundation of Liaoning Province of P.R.China (31), the Open Foundation of State Key Laboratory of Vision and Auditory Information Processing (Beijing University, 53), the Natural Science Foundation of Dalian City of P. R. China (6J3JH), the Open Foundation of Jiangsu Province Key Laboratory for Computer Information Processing Technology (Soochow University, KJS6), the Open Foundation of Key Laboratory of Image Processing and Image Communication (Nanjing University of Posts and Communications, ZK514) 1. Ûã Ç ² 1169. ü ² ÒÁ û 1871 1. School of Computer and Information Technique, Liaoning Normal University, Dalian 1169. State Key Laboratory of Vision and Auditory Information Processing, Beijing University, Beijing 1871 DOI: 1.136/aas-7-48 [1] èñ àß àß àß Å¾ ǑÅÙî DCT Í (Í ) Þ []  ² DCT Ù ùþ à ß Đ æ èí Ä î ðùóæ MPEG4 Á Ç Å Ô ü [3 7] Ý ß À ¹ ë Í Ö ( ÇðÔ ³ ) Þ [8]  Åè ß Ì ßÍÅ MP3 ÃǑ ² èùå Â Å Ö ë Í Đ Úñ1)  ² îá ² Ù Ë ò) ß Ó Ã (p,ε,q,d) Á ½ «ò3) ù ÃǑÅ Đ ò4) á Ð þí
3 å ñè ¹ MP3 ß 49  Åè ó ó ß Ì ßü Ǒ 16 Ò Ǒ ÁÈ û Ä Í Ç î ¹ ó Þ ² ó Ë Ǒñ èñ Ë ² ó (DCT DWT) Ë ² ² Ǒ A = {a(i), i < Length} Length Ǒ a(i) {,1,,,( p 1)} ð½ i p ǑÄ Þ ³ äǒ W = {w(i,j), i < M, j < N} w(i,j) {,1} ãä ä ½ i ½ ½ j ä Ǒ F = {f(i), i < Lsyn} f(i) {,1} (Ǒ Å Ù Ô Ò ½ß +1, 1 1) Lsyn Ǒ Đ Ã ó ó Ä (³ ù).1 Ò ǑÅî ä ä èñ ³ Â Ô Ë Í è Í ù» ä ½è ç Ǒ Arnold ä W è ç ǑË Å W 1 W 1 = {w 1 (i,j), i < M, j < N} Ǒ èâô¹ Ð þí ûä Í Ý Ù ß ² ½ Ò ¾ Þ Đ ĐǑ L 1 L û Ç ² ½ i Ä Ǒ A(i) = {a(il + k), k < L}( i < Length / L ) Ô L = L 1 + L, L 1 = Lsyn n, L = M N H+1, H Ǒ n Ǒ ( ü Ǒ 5) Ǒ A (ĐǑ L 1 L Ǒ A 1 A ) à ² Ä. ǑÅ Ý Đ Í (n ) àß ² Ç ÆùǑñ 1) A 1 Í Đ Lsyn Lsyn Þ è PA 1(m) õ n PA 1 (m) = { pa 1 (m)(i) = a 1 (i + m n), i < n, m < Lsyn} ) Ç PA 1(m) PA 1(m) = 1 n 1 pa n 1(m)(i),( m < Lsyn) i= 3) Þàß Þè PA 1(m) Í PA 1(m) è Í Ǒ pa 1 (m)(i) = pa 1(m)(i) + (PA 1 (m) PA 1(m)) (m)(i), i < n} Ǒ Í 1 = {pa 1 (m)(i), i < n} Ǒ Í PA 1 (m) = {pa 1 PA PA 1 ( (m) = IQ(PA 1 (m)) S 1 + S 1/,  Q(PA 1 (m)) = f(m) IQ(PA 1 (m)) S1 S 1/,  Q(PA 1 (m)) f(m) IQ(PA 1 (m)) = PA 1 (m). S 1 Q(PA 1 (m)) = mod(iq(pa 1 (m)),) mod Ǒ S 1 ǑÞĐ.3 ² 1) ( ) DWT Þ A ½ H û A H,DH,DH 1,,D 1 A H Ǒ A ½ H Õ D 1,,D H 1,D ÇǑ H A ½ 1 H ) Õ DCT Đ Õ ÍÅ ¹ä Ôü H Õ A H ½ DCT A HC = DCT(A H ) = {a (t) HC, t < L / H } 3) ² Ǒ Ý Đ ¹ Þ ² DCT ÞÆë M N DCT A HC A HC = { a (t)hc, t < M N a (t) HC, M N t < L / H
5 Þ 33 Þ Ã ñ a (t) HC = ( IQ(a (t) HC ) S + S /, ÂQ(a (t) HC ) = w 1(i, j) IQ(a (t) HC ) S S /, ÂQ(a (t) HC ) w 1(i, j) t = (i 1) N + j i < M j < N S ǑÞĐ j IQ(a (t) HC ) = (a (t) HC ) k S Q(a (t) HC ) = mod(iq(a (t) HC ),) Ô mod ð 4) DCT A HC DCT A H = IDCT(A HC ) 5) DWT A H ãç A H û ½ H DWT õ ² A 3 ÞĐ ó ü ÙÞ à Þ Đ ü æ³ ǑÞĐ Ö ç ³ ÞĐ Đ Í ( îǒ Æð ä) ü ùþđ û Đ ü ¹ Ý à [4] À Å Þ è ÞĐ ÂÞĐ Ã ÃǑ þ ò ÂÞĐ Ã ïá ÃǑ ½Æ à ÞĐ Í ³  ǑÅ Þ Đ Ú îó ß Ë ( Þ ùè ÞĐ Ç ) î Ë ËÆ ùþđ S 1 Ë ËÆ ùþđ S î ¹ ù ÞĐ S 1 ( Ù ) S 1 = log sa ( Ǒå ø ), s a = A s a = A Ä A ( ô ) ¹ ó ùþđ S ( Ù ² ) ÆùǑ 1) A (A õþ ) DWT A ½ H û A H,D H,D H 1,,D 1 A H Ǒ A ½ H Õ D H,DH 1,,D ÇǑ 1 A ½ 1 H ) Õ DCT Ôü H Õ A H ½ DCT A HC = DCT(A H ) = {a (t)hc, t < L / H } 3) ÞĐ S ó ü Ù ² DCT M N Ë ǑÅ Â Ä ÍÅ Ç ÞĐ ËÆû DCT M N ó ùþđ S S = log s b ( Ǒå ø ), s b = ÂHC = L / H 1 t=m N a (t) HC M N 4  ó ó ß Ù ß Ù ² î Á ½Æ 4.1 Ù Ô Ù ð ÙÁí ( Ǒ L 1 ) Ëð õ ² ùǒñ 1) Í ½ 3 ù ó ù Þ Đ S 1 ) Í ½. ù ÇÇ ( Áí ) n m PA (m) PA (m) PA (m) = {pa (m)(i) = a (i + m n), i < n, m < Lsyn} 3)  ² F  Ǒ F = {f (m) = PA (m). S1 mod, m < Lsyn} 4) ǑÅ í ÝæÙµ Å ² Ú ³ àß ù   F F ÃǑ Ù ² 4. ² Â Ô ²  ð Ð (¹Ì Ǒ A ĐǑ L ) Ë ² ùǒñ
3 å ñè ¹ MP3 ß 51 1) Í ½ 3 ù A ËÆ ó ù ÞĐ S ) A ½ H DWT A HC H H 1,D,D,,D A H 3) Õ A H DCT H = DCT(A ) = {a (t) HC, t < / L H} 4)  ²  Ǒ W = w (i,j) = ŵ (i,j) = a (t) HC / S 1 mod Ô w r(i,j) W r, ŵ (i,j) Ŵ 5  { 1 Âwr (i,j) è ( õè ) Ǒ 1 à ǑÅ Ý ß Ô ½ÅÙ Íý þíý û Þ [8] ß ½Å ³ ü ½Æð Ǒ 44.1 khz Ǒ 16 ³ ĐǑ 9.75 ² Å Daubechies-1 ü Ǒ H = ûü Å ĐǑ 16 Ò 11111111111 Ǒ ² Ù ð ü  ² Æð Ç Ä 1 Å ß Þ [8] ß þ Í Â (íµ Â ä èþ ³ NC ä BER ( i < M, j < N,t = (i 1) N + j) 5)  ² W ½ è ç ä W = {w (i,j), i < M, j < N} Á ð Ù ² Å ÙÄ Â ä Wr Á èí Ã Ç Ŵ Ä 1 þ ¹ Í Table 1 The watermark detection results for various attacks (r = 1,,,R) Ǒ (r = 1,,,R) ô ² ³ PSNR)  Åè ó ó ß Ì ßÍÅ Äü Ǒ 16 Ò Ǒ ÁÈ ËÆ ó ù ÞĐ Ç æ ² ó Ë Ë î å Ä Â ß Đ Í Þ á Ð Ò þ Ç» Ì ß Ù Ǒ MP3 þ Ö ¹ Í ( ðô ³ ) ß Ç ß Æð Ý îá ½Æ ÇÖÅ Ù MP3 Ï Ó éþ Þ Ô Ð Ð Ñ Mp3 Mp3 Mp3 Mp3 Mp3 Mp3.5kHz 11.5kHz 8kHz 1% 1% 3k 56k 11k 64k 48k 3k Þ NC 1..9996.3733.3773.37.9161.3851.3739 1. 1..9948.3619.3638.3667 [8] BER..4983.4939.4954.435.4833.4817.9.51.493.515 ß PSNR 37.5415 37.349 37.585.8633.8454 7.9397 13.9636.7441 37.318 37.317 36.8539 36.3864 35.3564 35.848 NC 1. 1. 1..93.8978.961 1. 1. 1. 1. 1..958.9748.934 ß BER.645.1414.443.73.144.57 PSNR 37.8174 37.4785 37.6815.8996.86 8.767 13.9789.7483 37.7747 37.9777 36.9516 36.5663 35.7669 35.871
5 Þ 33 References 1 Kim H J. Audio watermarking techniques. In: Proceedings of Pacific Rim Workshop on Digital Steganography, Kyushu Institute of Technology, Kitakyushu, Japan, July 3 4, 3 Lie W N, Chang L C. Robust and high-quality time-domain audio watermarking subject to psycho acoustic masking. In: Proceedings of the IEEE International Symposium on Circuits and Systems. Arizona, USA,. : 45 48 3 Yeo Kwon, Kim H J. Modified patchwork algorithm: A novel audio watermarking scheme. IEEE Transactions on Speech and Audio Processing, 3, 11(6): 381 386 4 Li Wei, Yuan Yi-Qun, Li Xiao-Qiang, Xue Xiang-Yang, Lu Pei-Zhong. Overview of digital audio watermarking. Journal on Communications, 5, 6(): 1 111 (Óå, è, Ó Ö, å, ö. Ò. ², 5, 6(): 1 111 ) 5 Xu C, Wu J, Sun Q, Xin K. Applications of digital watermarking technology in audio signals. Journal of the Audio Engineering Society, 1999, 47(1): 85 81 6 Swanson D., Zhu B., Tewfik H. Robust audio watermarking using perceptual masking. Signal Processing, 1998, 66(6): 337 355 7 Zhao Chun-Hui, Li Fu-Chang. A blind audio watermark algorithm based on DWT and DCT. Journal of Electronics & Information Technology, 3, 5(11): 157 1574 ( ËÞ, Ó. è DWT DCT ß. ², 3, 5(11): 157 1574) 8 Megías D, Herrera-joancomartí J, Minguillón J. A robust audio watermarking scheme based on MPEG 1 layer ¼ compression. Communications and Multimedia Security - CMS 3, Springer-Verlag, LNCS 963: 6 38. 3 å Ûã Ç ² íµ ² Ë ÜƲ Ò. ². E-mail: wxy37@63.net (WANG Xiang-Yang Professor. His research interest covers information security and multimedia processing. Corresponding author of this paper.) Ä Ûã Ç ² íµ Ü Æ Ç. (YANG Hong-Ying Associate professor. Her research interest covers multimedia technique and CAI.) Ûã Ç ² µ íµ². (ZHAO Hong Master student at Liaoning Normal University. Her research interest covers information security and digital watermarking.)