Efficient watermarking algorithm for digital audio/speech signal

Ensuring conflicting requirements such as imperceptibility, payload capacity, and robustness, becomes a great challenge for a robust audio/speech watermarking algorithm. To achieve this, this paper proposes a blind and robust audio/speech watermarking algorithm that combines the discrete Tchebichef moment transform (DTMT), the chaotic system of the mixed linear-nonlinear coupled map lattices (MLNCML), and discrete wavelet transform (DWT). The watermark is encrypted by the MLNCML system, and then it is embedded in the norm of the Tchebichef moments of the low frequency components. DTMT and MLNCML are used to achieve high robustness, high payload capacity, and high security, while DWT is used to achieve a satisfactory level of imperceptibility. In addition, the adopted strategy has a blind nature, where no original audio/speech is needed in watermark extraction. Compared to other existing audio/speech watermarking algorithms, the proposed algorithm gives better results in terms of robustness and payload capacity while keeping the embedding effect non-perceptible and undetectable. (c) 2021 Elsevier Inc. All rights reserved.

Automated summary

This paper proposes a robust audio/speech watermarking algorithm that combines multiple transformation methods to achieve high robustness, high payload capacity, and high security, with a blind extraction strategy.