Audio watermarking techniques using singular value decomposition

In an increasingly digital world, proving ownership of files is more and more difficult. For audio files, many schemes have been put into place to attempt to protect the rights of the digital content owners. In general, these techniques fall under the classification of Digital Rights Management (DRM). Audio watermarking is one of the less invasive schemes which embeds security into the data itself instead of in an outside layer meant to encapsulate and protect the data. There are many domains in which an audio watermark can be applied. The simplest is that of the time domain; often, however, other domains may be more desirable due to greater imperceptibility and robustness to attack. Common domains include the frequency domain, or domains similar to frequency through functions such as the Wavelet Transform. One domain of particular interest is that of the Singular Value Decomposition. The goal of this thesis is to propose and test many different watermarking schemes as well as test an existing watermarking scheme operating in the SVD domain in order to assess the viability of the SVD as a watermarking carrier domain. Different carrier matrices as well as bit embedding methods are explored. The use of a standard set of audio files was used to help test the systems; a standard set of watermarking tests was unavailable, so a comparable test bed was implemented and utilized.