Digital audio watermarking based on statistical inference

摘要

PURPOSE: A method for watermarking digital audio using statistical estimation is provided to prevent a watermark from being damaged by compression, filtering, DA-AD conversion and requantization and effectively use discrete cosine transform or discrete wavelet transform to detect the watermark rapidly. CONSTITUTION: An original audio signal is divided into block data having the same size and the block data is discrete-cosine-transformed or piecewise-constant-discrete- wavelet-transformed. Two index sets that are arbitrarily extracted are decided. DCT or PCDWT is applied according to a block data value given to the two index sets. The block data is inverse-discrete-cosine-transformed or inverse-piecewise-constant- discrete-wavelet-transformed to obtain an audio block into which a watermark bit being inserted.