A Lattice-Based Embedding Method for Reversible Audio Watermarking

TL;DR

This paper introduces a novel lattice-based reversible audio watermarking scheme called Meet-in-the-Middle Embedding (MME), which improves imperceptibility and robustness by scaling and adding lattice quantization errors, demonstrating superior performance in various metrics.

Contribution

The paper presents a new RAW scheme based on lattices that enhances imperceptibility and robustness through a novel error scaling and addition method.

Findings

MME achieves high SWR, ODG, and low BER.

Lattice quantization errors improve watermarking performance.

The scheme outperforms existing methods in simulations.

Abstract

Reversible audio watermarking (RAW) is a promising technique in various applications. To simultaneously meet the demand of achieving high imperceptibility and robustness, this paper proposes a novel RAW scheme based on lattices. The scheme is referred to as Meet-in-the-Middle Embedding (MME), in which the lattice quantization errors are properly scaled and added back to the quantized host signals. Simulations show that MME excels in a wide range of metrics including signal-to-watermark ratio (SWR), objective difference grade (ODG), and bit error rate (BER).