Abstract: An audio watermark algorithm based on odd-even quantification is proposed which has the features as follows.（1）The synchronous signal area is a section of audio signal which bares a definite length and a large amount of energy.So it is retrieved efficiently to satisfy the condition in the time domain of the audio signal.（2）The watermark image is divided into several blocks and then the small blocks and their serial number are embedded into the wavelet transformation low-frequency coefficients circularly to ensure that the watermark can be exacted accurately from the other audio signal session when some audio signals are cut off.（3）The watermark algorithm is a blind watermark detection algorithm which has no use to reference the original audio signal. The experiments prove the imperceptibility of the watermark and the robustness to the familiar attacks especially to the synchronous attacks and dithering attacks.

摘要： 提出一种基于奇偶量化的音频水印算法，算法具有如下特点：（1）在音频信号的时间域上检索满足条件的同步信号区，同步信号区是一段具有一定长度且能量较大的音频信号；这种时间域上的同步检索具有较高的效率；（2）水印图像划分成若干块，把每一个水印分块及其分块编号嵌入在一个同步信号区的小波变换低频系数上，水印图像划分成分块循环嵌入到检索到的同步信号区，当含水印的音频信号被裁剪后，可以确保在其他音频段提取出正确的水印分块；（3）水印检测不需要参考原始音频信号，是一种盲检测水印算法。仿真实验证明了算法对常见的音频信号处理具有较好的鲁棒性和不可感知性，对同步攻击、抖动攻击具有很好的鲁棒性。