Intelligent Decoding for Mean Quantization Based Audio Watermarking in the Wavelet Transform Domain

In this paper, a robust audio watermarking system, using mean quantization in the wavelet transform domain, has been proposed. Since the data is embedded in both the low and high frequency bands, selection of the correct result from these two bands is very important. In this paper, an intelligent decoder using two stage multi layer perceptron (MLP) neural network is proposed. Using this scheme, the attack is detected during the decoding process and the decoder is adapted to the same attack in order to extract the watermark data correctly. The simulation results show that using the intelligent decoder, in comparison to the previous scheme, the performance of detection after common attacks, such as lowpass, MP3 compression, highpass, echo, resampling, amplifying etc, is increased.