论文部分内容阅读
为了更好地提高音频数字水印的鲁棒性和不可感知性的平衡,利用神经网络的自学习和非线性映射能力,提出一种基于采样点倒置和反向传播神经网络的音频水印算法.算法首先对载体音频进行分帧,利用倒置的方法在音频中嵌入秘密信息;再利用BP算法对含密音频和载体音频建立人工神经网络,接受端利用该网络提取秘密信息.实验表明,该算法不仅具有良好的不可感知性,能够抵抗诸如低通滤波、加噪声、回声、重采样、MP3解压缩等攻击,特别还能抵抗DA/AD转换和各种去同步攻击.
In order to improve the robustness and imperceptible balance of audio digital watermarking, an audio watermarking algorithm based on neural network with self-learning and non-linear mapping ability is proposed. Firstly, the carrier audio is divided into frames, and the secret information is embedded in the audio by using the inversion method. Then the artificial neural network with dense audio and carrier audio is established by BP algorithm, and the receiver uses the network to extract the secret information. Experiments show that the algorithm not only It has good imperceptibility and is immune to attacks such as low-pass filtering, noise, echo, resampling, MP3 decompression, and especially to DA / AD conversion and various off-sync attacks.