Audio Watermark for Copyright Protection

Abstract

Audio watermarking is done by embedding side information in digital files without greatly affecting quality of songs and it is difficult to be removed. Advanced Audio Coding (AAC) in MPEG-2 is a relatively new digital audio standard developed in 90's. Compared with MP3, AAC has good quality with similar bitrates. That is why AAC is becoming popular these years and is widely used in commercial products like Apple (iPhone/iPad), Sony (PlayStation 3), etc. Psychoacoustics masking techniques such as frequency masking and temporal masking are implemented. By calculating signal-to-noise ratio (SNR) with the MDCT coefficients in consecutive inter-frames in frequency domain, where and how many extra bits can be added in the encoding process can be found. As the SNR of added bits are lower than certain threshold curve, songs with watermarks will be inaudible. After finding where watermarks can be added and when the encoders will enter a non-uniform quantification stage, the MDCT coefficients are grouped into scale-factor bands to calculate how many bits are needed for Huffman coding by using 11 Huffman tables. Least significant bits of the data bitstreams in particular locations got from above calculations are changed according to what data people want to embed inside. It is found that the number of watermark available is fewer than expectation. Strong power signals can add more watermarks without greatly affecting the quality. There are problems in retrieving watermarks form the decoder in Huffman decoding processes.