Statistical characteristic-based robust audio watermarking for resolving playback speed modification

In the field of audio watermarking, few schemes can resolve the problem of playback speed modification effectively. On the basis of statistical characteristic of DFT domain, a novel robust audio watermarking scheme against playback speed modification is proposed in this paper. Firstly, the original audio signal is segmented and then DFT is performed on each audio frame. Secondly, the absolute average value of DFT coefficients of each audio frame is computed and audio frames with distinct absolute average values are selected as watermark embedding regions. Thirdly, for one selected audio frame, it is divided into many segments, and the DFT domain relations among three neighboring segments are calculated and used to embed watermark bits. Finally, the watermark bits are repeatedly embedded into other selected audio frames. Experimental results show that the proposed scheme has acceptable imperceptibility and robustness against common audio signal processing attacks, including MP3 compression, re-quantization, re-sampling, etc. Furthermore, the ability of resisting playback speed modification and amplitude variation is also satisfactory.