Audio Watermarking Algorithm Robust to TSM Based on Counter Propagation Neural Network

Abstract

A novel audio watermarking algorithm robust to TSM based on Counter Propagation Neural Network (CPN) is proposed. Utilizing the learning and self-adaptive capabilities of CPN and adaptively changing the length of segment, the relationship between the important characters of audio and watermark signals was learned by using the variance of low frequency wavelet coefficients with strong stability as the input of CPN, with the purpose of embedding watermark. Experimental results show that the algorithm is very robust to common audio signal processing and synchronization attacks, such as Time Scale Modification (TSM).