一种数字音频水印嵌入算法的研究

Research on a Digital Audio Watermark Embedding Algorithm

结合人耳听觉模型，利用人耳掩蔽特性，提出了一种基于小波分解和复倒谱变换的音频数字水印算法。该算法通过对原始语音信号进行三级小波分解，提取出小波近似分量，从近似分量中选取重要系数。利用复倒谱变换系数的不相关性，对重要系数进行复倒谱变换实现解相关运算，最终把水印嵌入到语音复倒谱域中，实现了在语音信号中嵌入二值图像，提高了水印嵌入算法的不可听性。实验证明该方法在抗常见的信号处理如加噪、滤波、重采样、有损压缩等方面具有较好的稳健性。

Based on human auditory model and audio masking effects, an audio digital watermarking algorithm based on wavelet and cepstrum transform is proposed. This algorithm extracts the wavelet application components and selects important coefficients from them by decomposing original audio into three-level wavelets. It transforms the important coefficients in cepstrum domain to decompose the correlation of them by using the irrelevance of the cepstrum coefficients. Finally the watermarks are embedded in the cepstrum domain and a binary image watermark is embedded into the audio signal, hence the improvement of the imperceptibility of watermarks. Experimental results show that this algorithm has better robustness in resisting common signal processing such as noise adding, filtering,resampling and compression.