Abstract
During these recent years, different digital audio media are available due to internet and audio processing techniques. This is beneficial to our daily life but it brings the problem of ownership protection. As a solution, audio watermarking technique has been quickly developed. It consists of embedding information into digital audio data. This paper proposes an audio watermarking scheme operating in frequency domain using Discrete Cosine transforms (DCT). We take the advantage of hiding the mark in the frequency domain to guarantee robustness. To increase robustness of our schemes, the watermark is refined by the Hamming error correcting code. We study the band used to hide the watermark bits according essentially to the effects of the MP3 compression on the watermarked audio signal. To assure watermark embedding and extraction, we use neural network architecture. Experimental results show that the watermark is imperceptible and the algorithm is robust to many attacks like MP3 compression with several compression rates and various Stirmark attacks. Furthermore, the watermark can be blind extracted without needing the original audio.