An artificial neural network-based scheme for robust watermarking of audio signals

Digital watermarking is a technique by which the author information is embedded into the host signal (text, audio, image or video) that is imperceptible to the human senses. In this paper, we present a novel scheme for watermarking on audio signals using artificial neural networks (ANNs). The ANN is used to estimate the watermark scaling factor (WSP) intelligently from the knowledge of host audio signal. The power spectrum of the watermark signal remains below the minimum masking threshold (MMT) of the host signal when these WSFs are used in the watermarking process. This not only ensures inaudibility of the watermark signal, but also improves the capacity and robustness of the watermarking process. Using one music signal, we have shown the robustness of the scheme under some attacks to the watermarked audio signal.