A Reversible and Imperceptible Acoustic Watermarking Using Partially-applied Huffman Lossless Compression

In this paper, a reversible and robust acoustic watermarking based on lossless compression is proposed. Stereo speech data is represented by 16 bit for each sampling point and then divided into frames. Then Huffman lossless compression algorithm is applied to insignificant 4 bits in each sampling point partiality to reserve hiding capacity. An average of 0.7035 and the best 0.686 compression ratio depending on different frame lengths are achieved, which promises about 1.188 bits for payload hiding in each sampling point. Since Huffman algorithm is applied partially to each sampling point, stego data is comprehensive after embedding and complexity towards attack is promising. Result of Perceptual Evaluation of Speech Quality based on ITU-T recommendation P.862 and signal-noise ratio (SNR) show the proposed method achieved imperceptibility.