Abstract: Watermarking embedding algorithm based on auditory masking is proposed in this paper，which embeds watermarking based on masking threshold in low-band coefficients of DCT transformation.Calculating masking threshold of every critical band by auditory masking model，then quantization steps are controlled by masking threshold and value of audio carrier of the embedding points.At the same time，for meeting the require of fast embedding，we can find the best balance between time of embedding and robustness by algorithm proposed in this paper，and get the best repeating times of embedding watermark in one frame.Results of the experiment show that the embedded watermark are inaudible and have good robustness against kinds of attacks such as white noise，low-pass filtering，re-sampling and compression.

Key words: audio watermarking, masking model, quantization, robustness

摘要： 文中提出一种基于听觉掩蔽效应的音频数字水印算法，根据掩蔽阈值在DCT低频系数中嵌入水印。依据听觉掩蔽模型计算每个临界频带的掩蔽阈值，再根据水印嵌入点处的掩蔽阈值和音频载体值的大小共同控制量化步长。同时为满足水印快速嵌入的要求，通过文中提出的算法可以寻找到嵌入时间（计算量）与鲁棒性的最佳平衡点，得出一帧语音中嵌入水印的最佳重复次数。实验结果证明所嵌入的水印具有很好的不可感知性，对加白噪声、低通滤波、重采样和压缩等各种具有良好的鲁棒性。

关键词: 音频水印, 掩蔽模型, 量化, 鲁棒性