模糊RBF神经网络在音频零水印中的应用

The Application of Fuzzy RBF Neural Network in Audio Zero-Watermarking

针对当前零水印不能"嵌入"有意义水印的不足,构建了在小波域中基于神经网络的零水印系统,提出了一种基于模糊RBF神经网络的音频零水印方案,有效解决了音频水印的鲁棒性与透明性之间的矛盾.模糊神经网络模糊系统的隶属度函数和推理规则决定RBF神经网络的结构和学习算法.因为水印方案不改变原始音频数据,所以具有良好的透明性,实验结果表明,方案具有很强的鲁棒性.

Aiming at the shortcoming that now meaningful watermark can not be embedded into zero- watermarking, we proposed a zero-watermarking scheme based on neural network in wavelet field. A zero- watermarking scheme system based on fuzzy RBF neural network was constructed. It solves the contradiction efficiently between robustness and invisibility of audio watermark. For the network, the membership function and reasoning rules of fuzzy system are determined for special structure and learning algorithm of RBF network. Since the scheme did not change the original audio data, it had a good transparent. The experiment shows this scheme has strong robustness.