Framework for Hiding Information in Audio Sub-signals by Using Singular Spectrum Analysis with Psychoacoustic Model

Abstract

This paper proposes a method for hiding information in audio signals by using singular spectrum analysis (SSA) with a psychoacoustic model to enhance the sound quality of a watermarked signal from our previous work. In this framework, an original signal is decomposed into many sub-signals by applying the combination of the SSA-based method and the psychoacoustic model. The sub-signals to be embedded are chosen using signal analysis based on human perception. The watermark bits are embedded into selected sub-signals by modifying some parts of their singular spectra according to an embedding rule. Then, the accuracy of extracted watermark bits and the sound quality of the watermarked signal, compared with that of the original one, are used to verify the performance of our framework. The experimental results show that the proposed framework satisfies the evaluation criteria in terms of inaudibility and robustness. However, it has a significant trade-off between them.