Supplementary Schemes to Enhance the Performance of DWT-RDM-Based Blind Audio Watermarking

Abstract

This paper reports two schemes aimed at enhancing the performance of an algorithm based on rational dither modulation (RDM) for the blind watermarking of audio signals. The enhanced algorithm operates in a 4th-level approximation subband, which is obtained through discrete wavelet transform (DWT) decomposition. The first scheme involves the application of a two-tap FIR lowpass filter to mask the noise induced by watermarking, while the second scheme is meant to compensate for distortion introduced by vector modulation. These two schemes make it possible for the DWT-RDM algorithm to formulate audio watermarking with a payload capacity of 689 bits per second. The embedding process is organized as parametric vector operations, which allow the implantation of binary bits into DWT coefficient vectors using two variables, respectively, corresponding to lowpass filtering and distortion compensation. Experiments aimed at evaluating the quality of the resulting audio files indicated an improvement in average ODG score from $$-0.33$$-0.33 to $$-0.26$$-0.26. Furthermore, we observed a noticeable reduction in the bit error rates when conducting robustness tests against attacks, such as brumm addition, zero crossing, noise corruption, echo addition, MPEG-1 layer 3 compression, and time shifting.