一种基于线性预测系数多级矢量量化的语音零水印

Audio Zero Watermark Based on Multi-level Vector Quantization of Linear Predictive Coding Coefficient

提出一种基于线性预测系数多级矢量量化的数字零水印算法.通过对语音信号的线性预测系数进行多级矢量量化与待嵌入水印进行异或得到检测密钥,并在接收端进行盲水印提取.实验结果证明,该算法对于各种攻击尤其对幅度缩放、低通滤波和音量标准化具有较好的鲁棒性和不可见性.

An audio zero watermark based on the Multi-level Vector Quantization of linear predictive coding coefficient method was proposed.In the proposed method,the key with the method that the robust watermark was embedded in the Multi-level Vector Quantization of linear predictive coding coefficient which is picked up from the audio and index constrained respectively.Simulation results demonstrated that the effectiveness of the proposed algorithm in terms of robustness and imperceptibility was excellent especially in amplitude scaling,low-pass filter and volume normalization.