Robust Multiplicative Patchwork Method for audio watermarking

A multiplicative patchwork method (MPM) for audio watermarking is proposed in this paper. In order to embed watermark data within the host signal, two subsets of the host signal features are selected using two secret keys. Then, watermark data is inserted simply by multiplying one subset and leaving the other one unchanged. In order to have an error free detection in attack-free case, embedding is performed in the selected frames of the host signal which satisfies a certain condition. This method is implemented in the wavelet domain and approximation coefficients are used for data embedding. Furthermore, the inaudibility of watermark insertion is controlled using iterative approach aided by perceptual evaluation of audio quality (PEAQ) algorithm. Probability of error is derived and the validity of the analytical derivation is verified by simulation. Experimental results show that MPM is more robust than the previous patchwork methods against common attacks such as noise, filtering, MP3 compression, etc.