摘要
研究数字水印提高抗击性能,由于音频水印容量有限,同时盲检测方案中抗攻击能力普遍较弱,针对上述问题提出了一种在小波域中利用神经网络检测的数字语音水印算法。在数字语音载体的重要小波系数中隐藏了一幅不可感知的二值数字图像,通过二值数字图像中附加的模板对神经网络进行训练,经过训练后的神经网络几乎能够完全恢复嵌入到数字语音中的水印数据。仿真结果表明,小波域水印容量得到提高,算法对抗高斯噪声攻击、压缩、重采样等攻击,具有较强的鲁棒性,可以为实际应用提供依据。
To enhance the anti-attack ability of blind speech watermarking,this paper presents an algorithm based on neural network and wavelet transform.The watermarking technique hides an imperceptible image into a speech,then neural network is used to learn the relations between the speech and the watermark.The trained artificial neural network restores almost exactly the watermark imbedded in the digital speech.The experimental results show that this algorithm is robust and the watermark detection does not need original image carrier.In addition,the algorithm can be used in the audio copy right protection.
出处
《计算机仿真》
CSCD
北大核心
2011年第7期128-131,共4页
Computer Simulation
关键词
数字水印
小波变换
神经网络
信息隐藏
Digital watermark
Wavelet transform
Neural network
Information hiding
