A Novel Audio Watermarking Technique Using Data Dependent Decomposition and Quantization

Abstract

In this age of internet and with the advent of digital systems, audio, video and other documents are available in digital format. Internet has become a highway for digital data transportation. Many software’s and digital signal processing tools have been developed and distributed through the World Wide Web, thus providing tools to unauthorized personnel who make perfect copies of digital material. This may lead to large-scale unauthorized copying which might harm the software, film, book, and music industries. These threats over protecting copyright have lead to significant research in finding methods to hide copyright messages and serial numbers into digital media. These can help to identify copyright violators, and help to prosecute them. Watermarking can provide security of data depending on the type of application. For example in applications where full security is not needed watermark can be inserted to provide content monitoring and in-band signalling. Some characteristics of watermarking are robustness, Imperceptibility and capacity. User can work to improve any characteristic or all of three depending on the application of watermark. In the proposed algorithm, robustness, imperceptibility and capacity are targeted. A new method called empirical mode decomposition is used to decompose the audio signal. Quantization index modulation is used to embed the watermark. This method is a hybrid technique which utilizes both time and frequency information. EMD is adaptive or fully data driven unlike some other fixed kernel based methods. Quantization index modulation utilizes an array of quantizers to embed data in the audio samples. In this thesis QIM is varied by making it data dependent or adaptive. Along with literature survey some concepts on non stationary signal analysis have been introduced. Because audio signal is a non stationary signal. Also some techniques to quantify and analyse non stationary signals have been introduced thus providing a base for research work. Proposed algorithm has been simulated in MATLAB. The results are given with discussions. The proposed algorithm is also compared with existing methods and numerical values are compared. From the results it has been found that the proposed method has high capacity and capacity is compared with existing methods in the results section. Also the proposed is fairly robust to most of the attacks utilized during simulation process.