Research On Algorithm Of Audio Digital Watermarking

With the rapid growth of computers, network and multimedia technologies, powerful and low cost computers allow people to easily create and copy digital products of multimedia. The Internet has made it easy to distribute the information at very low cost. However, these technologies also can be illegally copied, modified and redistributed multimedia data without regarding for intellectual property rights ownership.With the mass use of the digital audio, the problem of protecting digital audio multimedia data becomes more and more important. The audio digital watermarking offers a solution to copyright violation problems, for it allows content creators to embed signature or copyright information into original signal, such as images, texts, sequences. All this addition is invisible or inaudible to the human, and they should also be statistically undetectable, and resistant to any attempts to attack it. The digital audio watermarking efficiently protects intellectual property of creators. Because of this, the digital audio watermarking is gradually becoming an important research direction in the digital watermarking. Audio watermarking algorithms based on non-compression domain and compression domain is conducted and proposed along with the author's viewpoints in this paper. The main work covers:1. The related knowledge of audio watermarking technique is first introduced; then the typical watermarking algorithms for both uncompressed and compressed audio are introduced and analyzed; finally, the evaluation framework for robust audio watermarking is also discussed.2. In non-compression domain, a novel algorithm for robust audio watermarking based on quantization in DCT domain is proposed. We design watermark algorithm according to DC and AC DCT coefficient play different influence in robust and inaudibility to tradeoff robust and inaudibility better. Firstly, we decompose the audio into no overlapping blocks every block contain eight data, then, every block is performed the DCT transform; watermark is embedded by modulating DC DCT coefficient or 4th AC DCT coefficient by turns, every block is embedded one bit message. Experimental results show watermark is robust to common digital processing operations such as add noise, loss compression, re-quantify etc, and inaudibility is satisfaction. To validate the stability of algorithm we also choose four types of music as carrier signal, experimental results show the stability of algorithm is satisfaction. Compares with the same type watermark algorithms, no matter robustness nor the audibility, this article's algorithm has the superiority.3. In compression domain, firstly the encoding method and the frame structure of MP3 are expounded in detail, then two watermark algorithms are proposed separately based on scale factor and the Huffman data in MP3 frame. By introducing the step of frame altering analysis, the former algorithm can realize watermarks embedded adaptively. By modifying threshold SNR (the ratio of signal to noise), algorithm can meet different requirements of someone system, that is, threshold SNR enables the algorithm to have a bigger flexibility. Watermarks also can be embedded with self-synchronization at expense of some embedding positions. During the algorithm which based on Huffman data, we define the parameter of code-couple altering as F , and determine the embedding positions according to F , the lower the F is, indicating the corresponding Huffman code-couple less sensitive, thus fit as a watermark embedding position,experimental results show both of the two algorithms can get a good audio effect with various types of audio.