Study On Robust Audio Watermaking

With the rapid development of computer network and multimedia technology, it is very convenient to distribute the digital audio. At the same time, it is very easy to pirate audio products because of the wide use of all kinds of audio software. So protecting audio becomes a more and more important issue today. As an effective technology, the digital watermarking has been paid an attention.The robust watermarking is studied in this paper, and the main work includes the following respects:(1) The basic theoretical knowledge of digital audio watermarking is introduced. Then the theory of the digital audio watermarking, the typical digital audio watermarking schemes, common attack methods and evaluation criterions are elaborated.(2) The basic principals of the psychoacoustic model, the complex cepstrum analysis and filter banks are introduced. An adaptive audio watermarking algorithm based on filter banks is proposed. Each audio frame was divided into32sub-band signals. Then watermarks were embedded in the entire frequency domains by combining the complex cepstrum transform. This scheme adjusts the embedding strength of watermark by the psychoacoustic system to enhance the watermark robustness, and embedding watermark in sub-band domain makes the scheme be combined with the human auditory characteristic, which enhances the imperceptibility. Simulation results show that the proposed algorithm is robust against some common attacks such as adding noise, low-pass filtering, re-quantization, MPEG-1Layer III (MP3) compression. The scheme realizes a better balance between the robustness and imperceptibility.(3) The basic principle and characteristics of SVD and DWT are introduced. DWT has good multi-resolution analysis properties, and the singular value also has a very good stability. So an audio watermarking algorithm is proposed, which based on wavelet transform and singular value. The two methods are applied in the embedding watermarking. At the same time, the synchronization code and number is used against the synchronization attack. Simulation results show that this algorithm is robust against some common attacks such as adding noise, re-sampling, MP3compression, cropping.