Audio Watermarking Detection Resistant to Time and Pitch Scale Modification

Abstract

In this paper, an efficient audio watermarking detection is developed to revive a watermark attacked by hazardous synchronization attacks, such as time-scale and pitch-scale modifications (TSM and PSM). It outperforms the robustness of our previously implemented audio watermarking scheme by resisting excessive distortion of up to ±10%. The key idea of the method is adaptively aligning the frequency spectra that have been scaled by the attack, so that the synchronization positions can be retrieved for recovering the embedded watermark. Without resorting to the host audio signal, the detection procedure will automatically find a matched state and obtain the highest detection accuracy. Experimental results of robustness tests have proved that the proposed detection method can efficiently handle the severe TSM and PSM attacks and successfully extract the watermark.