Abstract
Being reversible, the watermarking information embedded in audio signals can be extracted while the original audio data can achieve lossless recovery. Currently, the few reversible audio watermarking algorithms are confronted with following problems: relatively low SNR (signal-to-noise) of embedded audio; a large amount of auxiliary embedded location information; and the absence of accurate capacity control capability. In this paper, we present a novel reversible audio watermarking scheme based on improved prediction error expansion and histogram shifting. First, we use differential evolution algorithm to optimize prediction coefficients and then apply prediction error expansion to output stego data. Second, in order to reduce location map bits length, we introduced histogram shifting scheme. Meanwhile, the prediction error modification threshold according to a given embedding capacity can be computed by our proposed scheme. Experiments show that this algorithm improves the SNR of embedded audio signals and embedding capacity, drastically reduces location map bits length, and enhances capacity control capability.
Highlights
Reversible watermarking is a technique that means after watermark information was embedded into the host data, original data can be restored while watermark information is extracted through a series of processes
We proposed a novel reversible audio watermarking scheme based on improved prediction error expansion and histogram shifting which combined with the characteristics of digital audio data
We can see that the linear prediction coefficients generated by Differential evolution (DE) algorithm achieve better SNR and capacity than fixed coefficients for all the types of test audios in the same condition
Summary
Reversible watermarking is a technique that means after watermark information was embedded into the host data, original data can be restored while watermark information is extracted through a series of processes. Tian [7] proposed an image reversible watermarking algorithm using difference expansion technique. In [10], Yan and Wang proposed a reversible audio watermarking scheme based on prediction error expansion method. Watermarking information was embedded by expanding the difference between the prediction and current amplitude This algorithm mapped the audio data to integer interval and achieved recovery of original audio signals. In [11], a reversible audio data hiding algorithm based on linear prediction and error expansion was present by A. We proposed a novel reversible audio watermarking scheme based on improved prediction error expansion and histogram shifting which combined with the characteristics of digital audio data. According to the characteristics of audio data, we introduced histogram shifting technique which can dramatically reduce the embedding location information.
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have