Synchronized Blind Audio Watermarking via Multilevel DWT and Windowed Vector Modulation

A method emerging from multilevel DWT is introduced to perform synchronous blind audio watermarking. While the repetitive pattern hidden in the approximation subband is in charge of frame synchronization, the wideband signal covering the to detail subbands is selected to hide binary information. By tracing zero-crossings of the synchronous signal, we demonstrate how to recalibrate frame positions and implement watermarking on a RDM framework. Furthermore, a windowed vector modulation scheme in DCT domain is developed to enhance watermark imperceptivity. Experimental results show that the embedded watermarks hold sufficient robustness against commonly encountered attacks. The proposed synchronized watermarking method currently attains a payload capacity of 86 bits per second. For the desynchronous attacks such as resampling and cropping, the proposed technique effectively retrieves the watermark bits with excellent accuracy.