The design of optimum blind audio watermark decoder using higher order spectrum

Digital watermarking is an effective means to resolve copyright protection problem by embedding additional information into digital media. The errors of watermark detector can undermine the credibility of a watermarking system. It is crucial to minimize the false detection rate in watermark detection. Many proposed watermark detector were based on correlation detection. In these cases, the performance of the detector depends mainly on signal to noise ration. With decreasing of SNR, the detection performance will deteriorate. A new optimum blind audio watermark detector is proposed based on higher order spectrum of the audio based on discrete wavelet transform domain in this paper, as higher order statistics are insensitive to Gaussian noise no matter it is white or colored. Bispectrum of audio is used during detection with constant false alarm rate, which is very useful in signal detection. The advantage of detector based on bispectrum is proved in theory and the experimental results show that the detector based on bispectrum is more robust than the detector based on correlation.