Audio Watermarking by DWT-SVD-BFO

 The main aim of this work is to develop a new watermarking algorithm within an existing discrete wavelet Transform (DWT) and singular value decomposition (SVD) framework. This resulted in the development of a combination of DWT-SVD-BFO (bacterial foraging optimization) watermarking algorithm. In this new implementation, the embedding depth was generated dynamically thereby rendering it more difficult for an attacker to remove, and watermark information was embedded by manipulation of the spectral components in the spatial domain thereby reducing any audible distortion. Further improvements were attained when the embedding criteria was based on bin location comparison instead of magnitude, thereby rendering it more robust against those attacks that interfere with the spectral magnitudes. The further aim of this thesis is to analyze the algorithm from a different perspective

Project Introduction

Technological advances in computing, communications, consumer electronics and their convergence have resulted in phenomenal increases in the amount of digital content that is being generated, stored, distributed, and consumed. The term “content” broadly refers to any digital information, such as digital audio, video, graphics, animation, images, text, or any combinations of these types. This digital content can be easily accessed, perfectly copied, rapidly disseminated and massively shared without it losing quality, as opposed to the situation with earlier analogue media, such as audio cassettes and Video Home System (VHS) tapes.

However, these advantages of digital media formats over analogue transform into disadvantages with respect to copyright management, because the possibility of unlimited copying without a loss of fidelity has led to a considerable financial loss for copyright holders. In our work audio watermarking is the target area because of high digitally spread content of audio, music etc.

The algorithm process must possess some characteristics like imperceptibility, exact detection etc. To match these, our work will be evaluated on the basis of peak signal to noise ratio (PSNR) and normalized cross correlation (NCC). Till now many researchers have used optimization algorithms to tune the gain value but signal value for whole audio signal was used. This increases the chance of perception for gain value by third user. In our work, we have divided the audio signal into chunks and gain value is calculated by BFO as per the number of chunks. The number of chunks depends upon the watermark level used. We have developed the dynamic MATLAB script which can work for desired DWT levels.