Adaptive audio watermarking for Indian musical signals by GOS modification

Spread spectrum audio watermarking technique is developed by modifying the group of samples (GOS) in DCT domain. To check the robustness of the technique the watermarked signal is passed through different signal processing attacks and is observed that the watermark remains present after all signal processing attacks. The system is able to hide 1024 data bits in a clip of 6 sec MP3 song i. e. approximately 170 bits per second. The technique is also tested with the subjective listening test and it is confirmed that the there is no perceptible difference between the watermarked signal and the original host audio signal. It is also observed that the SNR is improved compared to the conventional spread spectrum technique.