International Journal of Computer Applications in Engineering Sciences [VOL I, ISSUE II, JUNE 2011] [ISSN: 2231-4946] 150 | Page Comparative Study of Audio Watermarking and a Scheme for an Efficient, Robust and Imperceptible Watermarking Kamini Pawar # , S. K. Parchandekar * , S. G. Tamhankar # # Department of Electronics, Walchand College of Engineering, Sangli, Maharashtra, India Department of Electronics, Walchand College of Engineering, Sangli, Maharashtra, India 1 kamini.pawar@gmail.com 3 sgtamhankar@yahoo.com 2 parchandekar@rediffmail.com Abstract— Digital audio watermarking is a technique for embedding additional data along with the audio signal. Embedded data is used for copyright owner identification. Digital audio watermarking techniques include traditional watermarking and zero-watermarking scheme. In this paper, traditional watermarking scheme is compared with zero-watermarking scheme. In traditional watermarking scheme, LSB technique is implemented in wavelet domain. Zero-watermarking scheme is based on steady sign of certain DWT-DCT co-efficients with maximum absolute value. The experimental result shows that the zero- watermarking scheme is more efficient, robust and imperceptible over traditional watermarking scheme. Keywords— Traditional watermarking, LSB technique, zero-watermarking, DWT, DCT. I. INTRODUCTION Digital watermarking is a process of embedding additional secret information into the digital signal such as audio, image or video signal [1]. The research on audio watermarking is not as mature as compared to image watermarking. Indeed only few publications on audio watermarking exist. Among various audio watermarking schemes, the LSB technique is a powerful method, as it enables high rate of information hiding and ease of implementation [2],[3]. But the robustness of this method is very low. The watermark bits can be easily detected and changed. Watermark extraction rate is also very low when watermarked signal is subjected to different types of attacks. Other traditional watermarking schemes based on different transforms are also present [4], [5]. In traditional watermarking techniques, either in spatial domain, transform domain or dual domain [6-9], the embedding of the watermark into the host audio inevitably introduces some audio quality degradation. Another problem is the inherent conflict between imperceptibility and robustness. Then zero- watermarking technique was proposed by some researchers to solve these problems [10-12]. Instead of embedding the watermark into the audio signal, the zero-watermarking approach just constructs a binary pattern based on the essential characteristics of the host audio signal and uses them for watermark recovery. This paper compares the traditional watermarking techniques with the zero-watermarking technique by implanting algorithms of both the techniques. Experimental results show that zero-watermarking technique is more robust than the traditional watermarking schemes and it reduces the trade-off between imperceptibility and robustness. An effective zero-watermarking scheme among different audio watermarking techniques is proposed. II. TRADITIONAL WATERMARKING - LSB TECHNIQUE While implementing LSB technique, the watermark is embedded in frequency coefficients. In this paper DWT is used to implement this technique. During embedding process, appropriate level of DWT decomposition is carried out and then every pixel of the binary image is embedded into the LSBs of the decomposed coarse frequency coefficients. During extraction process, IDWT is carried out and image pixels are retrieved from the LSBs of the coarse frequency coefficients. Embedding a single image watermark in an entire audio signal leads to unsuccessful attempt of tampering detection, if other part of the signal i.e. part of the audio signal where watermark is not embedded, is attacked or cropped. Hence, solution to this problem is segmentation of audio signal. Concept of segmentation- Audio signal is segmented into different parts or segments. Then some or all of the segments are chosen for watermark embedding. Copies of the watermark image are embedded in each selected segment of the audio. In this paper, results of different combinations of images and audio signals are obtained using LSB technique.