A Mean Quantization Watermarking Scheme for Audio Signals Using Singular-Value Decomposition

Recently, research in audio watermarking technology makes it a promising tool for copyright protection of digital audio files in networked environments. In this paper, we propose a new mean quantization based watermarking mechanism for audio based on the technique due to Singular Value Decomposition (SVD). In this method, the host signal is divided into a number of two-dimensional matrix frames. The SVD is then employed to every frame, and also the Euclidean norm of the Singular Values (SVs) are calculated for every frame. The watermark is then hidden into an audio signal by quantization of the norm of the SVs. The watermark is extracted blindly using the inverse process. Both the subjective and objective tests demonstrate good imperceptibility of the watermark in the audio file. Furthermore, the experimental results show good robustness of the proposed scheme against various common audio attacks and Stirmark benchmark attacks. The false-negative error of the method is very near to zero against Stirmark and audio attacks. The proposed scheme has a high payload, and its performance is superior compared to other related watermarking methods for audio signals. Finally, the proposed method achieves better trade-offs between conflicting requirements of imperceptibility, payload, and robustness.