High Resilience Diverse Domain Multilevel Audio Watermarking with Adaptive Threshold
This work addresses the need for more resilient audio watermarking against attacks, though it appears incremental as it builds on existing domain methods with novel detection techniques.
The paper tackles the problem of robust audio watermarking by proposing a dual-domain embedding scheme (DCT-SVD and DWT-SVD) and adaptive threshold detection algorithms (AOT and AOTx), resulting in improved subjective and objective quality and reduced susceptibility to signal processing attacks with minimized Bit Error Rate (BER).
A novel diverse domain (DCT-SVD & DWT-SVD) watermarking scheme is proposed in this paper. Here, the watermark is embedded simultaneously onto the two domains. It is shown that an audio signal watermarked using this scheme has better subjective and objective quality when compared with other watermarking schemes. Also proposed are two novel watermark detection algorithms viz., AOT (Adaptively Optimised Threshold) and AOTx (AOT eXtended). The fundamental idea behind both is finding an optimum threshold for detecting a known character embedded along with the actual watermarks in a known location, with the constraint that the Bit Error Rate (BER) is minimum. This optimum threshold is used for detecting the other characters in the watermarks. This approach is shown to make the watermarking scheme less susceptible to various signal processing attacks, thus making the watermarks more robust.