A dual transform audio watermarking algorithm

In this paper, a non-blind digital audio watermarking algorithm that satisfies the minimum requirements of optimal audio watermarking set by the International Federation of Photographic Industry (IFPI), is proposed. The algorithm does not degrade perception of audio, offers an SNR value of more than 44 dB, offers around 1387 bps data payload, and it is robust against common audio processing operations such as Gaussian noise addition, MP3 compression, re-quantization, re-sampling, echo, low-pass, high-pass, and band-pass filtering. The IFPI requirements were met by the proposed algorithm as it exploits the attractive properties of two powerful mathematical transforms; the Discrete Wavelet Transform (DWT), and the Singular Value Decomposition (SVD). DWT is applied to achieve robustness as it decomposes the original audio signal in such a way to scatter watermark bits throughout the signal. SVD provides imperceptibility to the proposed algorithm by embedding watermark bits onto the diagonal singular values of the S matrix produced by SVD. The effectiveness of the algorithm is demonstrated by a set of experiments using pop, instrumental, and speech audio scripts.