A Robust Audio Watermarking Algorithm Based on Statistical Characteristics and DWT+DCT Transforms

This paper considers quantization based audio watermarking algorithm based on both Discrete Wavelet Transforms (DWT) and Discrete Cosine Transform (DCT). The embedding information used is a binary image. Different techniques like Arnold Transform and Linear Feedback Shift Register are used for image encryption. Different wavelet filters are considered and the performance is evaluated. Watermark embedding is performed by mean quantization of DCT values of n^th level low frequency components where n ε {3, 4, 5}. Low frequency components contains majority of the signals energy, which helps provide robustness. Synchronization codes are embedded in the low frequency components of the audio signal, and it makes algorithm more robust against desynchronized attacks. Experimental results show that the proposed algorithm is inaudible and robust against common signal processing techniques. Our results are better than those obtained using techniques involving only single domain transformations such as [2], [9] and [10].