采用量化MCLT变换系数的音频数字水印盲检算法

提出了量化 MCLT 变换系数的音频数字水印盲检新算法。嵌入的水印为视觉可辨的二值标志图像,对水印降维处理后,利用密钥通过 PN 序列发生器产生的伪随机序列调制水印信息,然后通过量化 MCLT 系数的方法将其嵌入音频中。通过大量实验,得到了决定水印鲁棒性和不可感知性的量化参数选择上限。实验结果表明,该算法对所嵌水印的音频质量未见降低,对信号处理和随机裁剪型的攻击具有很高的鲁棒性。

An audio digital watermarking blind algorithm based on quantizing MCLT transformation coefficient

A new audio digital watermarking blind algorithm based on MCLT coefficient quantization is presented. The embedded watermark is a visually recognizable image. After converting the binary watermark image into one-dimension sequence, the sequence is then modulated with m-sequence. Finally, the watermark is embedded into the audio signal by modifying the selected coefficients in the quantification process. The upper limit of quantization parameter is found by lots of listening test. Experimental results demonstrate that the embedding strategy is effective and the embedded watermark is inaudible and sustainable against compression algorithms, and common signal processing manipulation attacks.