| The digital information has brought many changes to our life in recent decades. With the rapid growth of digital information, a lot of new techniques are coming into being. Digital watermarking is one of them. This thesis presents a semi-fragile audio watermarking scheme which can be applied to content authentication or copyright verification. The major contribution of this thesis is the introduction and implementation of a unified copyright verification and content authentication algorithm for audio signal.; In our approach, we embed the watermark in the discrete wavelet domain of an audio by using quantization technique. The discrete wavelet domain has both the spatial and frequency information which make it possible to detect various modifications. The advantage of using quantization technique is that we could extract the watermark without using original audio.; In the procedure of watermark extraction, some signal processing operations may change the length of testing audio. We employ a matching filter to locate the start point of the watermark. This filter can be generated during the embedding procedure. The cost of the filter is very small.; We conduct several experiments for evaluating the performance of this new watermarking technique on both of two applications---content authentication and copyright verification. Our experimental results show that this scheme is robust to mp3 compression, additive noise, and filtering attacks. At the same time, the embedded watermark can also be used to check whether the audio content has been modified or not. If the audio is determined to be modified, an assistant program can be invoked to find out where the modification is. |
