In this paper, a blind digital audio watermarking scheme is proposed. A psychoacoustic model is used to ensure the imperceptibility of the embedded watermark. The watermark is embedded in the discrete cosine transform (DCT) domain by changing the phase of five information carriers determined using the psychoacoustic model. Blind recovery is achieved through the application of a novel synchronization scheme, the peak point extraction (PPE) scheme. Error correcting codes are used to enhance the robustness of the system. Informal listening reveals excellent imperceptibility of the embedded watermark. The watermarked signal was put through various forms of attacks. Experimental results show that recovery of watermark is perfect for signals not subjected to attack and the system is robust against attacks such as cropping and re-sampling.
Foo Say Wei, Xue Feng, Li Mengyuan
2005 IEEE International Symposium on Circuits and Systems