基于声学心理模型的固定门限音频水印方法

An Audiowatermark Method Based on Psychoacoustic Model and Predefined Threshold

提出了一种基于声学心理模型和小波包变换的的音频水印算法,适用于立体声音频信号和单声道音频信号。对于立体声音频信号,在左右声道加入不同的水印信号;对于单声道音频信号,在奇偶帧加入不同的水印信号。左声道水印通过源信号包络与一个只有版权拥有者持有的随机序列加权生成水印信号,并且通过对得出信号进行小波包和声学心理模型分析控制水印信号的强度;右声道水印信号为幅度限制在可听门限的白噪声,用来控制检测门限。得到的水印信号在时域嵌入,水印的检测无需原文件。该方法能够将检测门限控制在事先设定的范围内,具有较好的隐秘性;嵌入的水印信号能够抵抗时域的切割、加噪攻击和一般的音频变换算法。

An audio watermark algorithm based on psychoacoustic model and wavelet packet analysis which is applicable both to stereo and mono audio is presented. Different watermark signals are embedded into the left and right channels as for the stereo or odd and even frames as for the mono-audio. Watermark signal of the left channel is produced by weighting the original signal with a random sequence possessed only by the copyright owner which will later be shaped by the wavelet package and psychoacoustic model analysis. Watermark signal of the right channel whose magnitude is limited below the audible threshold, is used to control the detection threshold. The obtained watermark signal is embedded in the time domain. The detection of the watermark does not require the original signal. This watermark algorithm can contain the detection threshold into a predefined scope and attain quite good imperceptibility. The embedded watermark can resist time cropping, noise attack and normal audio transform algorithm.