Robust Audio Watermarking Algorithm Based On Adaptive Quantized Mean Against Synchronization Attack

The audio signal is a function on the time axis.When the watermark audio is attacked by different synchronization signals,the synchronization of the audio signal and the original watermark signal on the time axis is destroyed.Most digital audio watermarking algorithms are designed on the premise that the location of watermark extraction is consistent with that of watermark embedding,so the effect of resisting synchronization attack is not good.When the same algorithm is used to embed the watermark into the host audio without and with silent clips,the robustness of the two algorithms is often different.This paper proposes a robust audio watermarking algorithm with adaptive quantization mean against synchronization attack.This algorithm uses the global mean feature of the host audio signal frame instead of the local feature to embed the watermark signal,so as to improve the robustness of the watermark system against synchronization attacks.This algorithm provides an effective solution to overcome the continuous offset problem caused by synchronization attack.Using the mean feature of host audio can prevent the disappearance of local feature points after synchronization operation.In the process of quantization embedding,the initial SNR of the original host audio signal and the watermark signal is used to select the adaptive quantization step size to overcome the problem of a large number of repeated experiments.Based on the analysis of the influence of the silent segment of the host audio on the overall effect of the audio watermarking system,this study proposes to cut and extract the silent segment and non-silent segment according to the characteristics of the host audio,and embed the watermark signal into the non-silent segment of the host audio to eliminate the influence of the silent segment of the host audio on the watermark embedding effect.