Watermarking power estimate and adaptive embedding algorithm for audio signal based on acoustic masking characteristic

A method of estimating the maximum of the audio watermarking is presented since watermarking power is an important role to improve the watermarking robustness. Using the auditory masking property and MPEG-2 auditory model, the method calculates the watermarking power in per-frame, and estimates the whole maximum watermarking power. It is also used to construct adaptive embedded watermarking models. The experiments show that the more power embedded in the watermarking the faster the rate of the change of watermarking robustness. According to the self-change characteristic of the audio signal, this adaptive embedded watermarking model can adjust the embedded intensity dynamically and improve the whole watermarking power without losing auditory quality. This method not only makes it easier to estimate watermarking power through experiments, but also reduces noise caused by fixed embedded coefficients, even with the same embedded watermarking power.