A NOVEL METHOD FOR REAL-TIME AUDIO WATERMARKING USING WAVELET TRANSFORM Mahmood Movassagh, Ali Asghar Beheshti Shirazi Department of Electrical Engineering, Iran University of Science and Technology (IUST) mmovassagh@ee.iust.ac.ir , abeheshti@iust.ac.ir ABSTRACT In this paper we propose a novel method for audio watermarking which is appropriate for real-time applications. Here the energy proportion of two successive low-middle frequency bands of each audio signal frame is computed and through the use of Index Modulation technique we embed the watermark bits into these proportion values. Low complexity is the most important characteristic of the proposed method making it appropriate for real –time applications. The method is a blind one, that is it does not need the original signal for extracting watermark bits. Experimental results have shown that it has a very good robustness against common watermarking attacks such as mp3 compression, low pass filtering, resampling, requantizing and adding noise. Index Terms- Audio Watermarking, Real-Time, Wavelet Transform, Quantization Index Modulation (QIM) 1. INTRODUCTION With the growth of multimedia production and broadcasting, security issues have been concerning for multimedia producers. In recent years watermarking has provided good means for protecting multimedia against illegal tasks especially for copy protection, ownership proof and modification protection. 978-1-4244-1643-1/08/$25.00 ©2008 IEEE Although most multimedia watermarking works have been dedicated to image watermarking so far, during the last decade many audio watermarking methods have been proposed [1]-[4]. For example, techniques based on perceptual masking, spread spectrum communications (SS), phase coding, phase modulation, echo hiding and so on [4, 5]. Generally we can divide watermarking techniques into two domains: Time domain and Frequency domain. In time domain methods watermark bits are embedded directly into the signal time samples and in the latter, after taking one of the usual transforms such as FFT, MDCT and WT from the signal, watermark bits are embedded into the resulting transform coefficients [2, 5, 6]. Among the existing transforms, wavelet transform has some advantages in audio signal processing. Its inherent frequency multiresolution and logarithmic decomposition of frequency bands resembles human perception of frequencies because it allows the decomposition to mimic the critical band structure of the human auditory system [7, 8]. The two other mentioned transforms, FT and MDCT, produce equal- spaced frequency bands which do not correspond to human perception of the frequencies and to get acceptable frequency resolution in low frequencies, high frequency resolution is over-detailed. Therefore, they are quite inefficient for audio processing compared with the wavelet transform. Using transform based methods provides a better perception quality and robustness against common attacks while causing more complexity. For real time applications complexity is the most important task that should be considered. In [12] an efficient audio watermarking 000083