An Efficient Audio Watermark Algorithm with Strong Robustness

An efficient audio watermarking algorithm with strong robustness is proposed in this paper. This algorithm firstly makes non-negative matrix factorization(NMF) transformation to host audio carrier, secondly does singular value decomposition (SVD) decomposition to the coefficient matrix, then the algorithm introduces hybrid computing model, embeds watermark into the singular value of the coefficient matrix. The watermark embedding intensity is adaptively adjusted by the way of mixed matrix coefficients; the algorithm introduces principal component analysis(PCA) method accelerates the convergence speed. The experiment shows that this algorithm has the strong robustness to additive noise, low-pass filter, re-sampling, damaging compression etc. The algorithm¿s convergence speed improves 80.38% comparing with random initialization method. The algorithm¿s embedded capacity is large. The algorithm is a blind watermarking scheme.