中文摘要：

根据原始水印构造匹配滤波器，并结合小波估计方法，提出脆弱音频水印的盲检测算法。在假定嵌入水印为噪声的情况下，利用小波不需要原始信号辅助信息即可估计该信号的特点，首先根据给定的估计门限估计了加入水印后的原始信号，从而实现了水印与音频信号分离。然后将水印与估计产生的噪声构成的混合信号送人构造的匹配滤波器作相关运算，根据滤波器的输出响应判断水印及载体是否受到攻击。实验结果表明，所提出方法对噪声、有损压缩、滤波、随机插入与删除等操作具有较高的敏感性。

英文摘要：

Constructing matched filtering based on initial watermarking and combining with wavelet estimation method, this paper presents a blind detect algorithm. Assuming the watermarking as noise, the proposed method adopts the Wavelet Packet 1-D algorithm to estimate the audio signal using a selected threshold criterion without original audio signal. The mixed separated noise ,which includes embedded watermarking and estimation error,is sent into the constructed matched filter. Whether the audio signal has been attacked or not are detected by using the impulse response of the matched filter. Experimental results show that the proposed approach has highly sensitivity about add noise, lossy compression,filtering and insert or delete samples randomly.