基于小波和倒谱域的数字音频水印算法

DIGITAL AUDIO WATERMARKING ALGORITHM BASED ON WAVELET AND CEPSTRUM DOMAIN

给出一种基于小波和倒谱域的强鲁棒数字音频水印算法。该算法利用人耳听觉模型和频率掩蔽效应,选择出音频小波变换后对人耳听觉不敏感的低频系数,对其进行分帧并作倒谱变换。再利用倒谱域能量集中在零点附近的特性,将一段伪随机序列嵌入在每帧相对稳定的倒谱系数中,从而实现水印的嵌入。采用相关检测的方法,根据相关值的极性来提取水印信息。实验表明,嵌入水印后的音频信号不仅具有良好的不可感知性,而且对诸如加噪、低通滤波、重采样、重量化、回声、MP3压缩和去同步等攻击具有很强的鲁棒性。

A strong robust digital audio watermarking algorithm based on wavelet transform and cepstrum domain is proposed in this paper.First,the audio wavelet-transformed imperceptible low frequency coefficients are selected using the human ear hearing model and the masking effect on the frequency,then these coefficients are frame-separated and are conducted the cepstrum transformation.According to the characteristic that the energy in cepstrum domain focuses on both sides of the centre,the pseudorandom sequence is embedded into relatively stable cepstrum coefficients in each segment so as to achieve watermark embedding.The watermarks are extracted out using the method of correlation detection according to the polarity of the correlated value.Experiments show that the audio signal with embedded watermarks has good imperceptibility and strong robustness against various common signal processing attacks,such as noise adding,low-pass filtering,resampling,requalisation,echo,MP3 compression and desynchronisation,etc.