Robust Audio Watermarking Using Graph-based Transform and Singular Value Decomposition

Abstract: Graph-based Transform (GT) has been recently leveraged successfully in the signal processing domain, specifically for compression purposes. In this paper, we employ the GBT, as well as the Singular Value Decomposition (SVD) with the goal to improve the robustness of audio watermarking against different attacks on the audio signals, such as noise and compression. Experimental results on the NOIZEUS speech database and MIR-1k music database clearly certify that the proposed GBT-SVD-based method is robust against the attacks. Moreover, the results exhibit a good quality after the embedding based on PSNR, PESQ, and STOI measures. Also, the payload for the proposed method is 800 and 1600 for speech and music signals, respectively which are higher than some robust watermarking methods such as DWT-SVD and DWT-DCT.