An error resilient scheme of digital watermarking for MP3 streaming audio

The recent massive growth of networked multimedia has caused problems relative to the protection of intellectual property rights. This is particularly true to MP3 audio data. These types of protection systems involve the use of both encryption and authentication techniques. In this thesis we describe a form of authentication known as digital watermarking. A novel scheme to embed digital watermark to MP3 music using Quantization Index Modulation (QIM) algorithm is proposed in the thesis. A pseudorandom key is generated as watermark message which is embedded into MDCT coefficients of MP3 directly. At the receiver, the MP3 is decoded partially and the watermark is extracted frame by frame. Furthermore we discuss the recovery method of the watermarking after the watermarked MP3 audio is streamed on the network. Distributing digital watermark on Internet through streaming audio is a challenging task because most digital watermarking algorithm is very sensitive to packet loss due to the associated synchronization problem. On the current Internet, packet loss is almost inevitable since its backbone protocols are operating in best-effort manner and do not guarantee the successful delivery of data packets. Therefore it is very essential to develop a scheme that resists the damage caused by packet loss to audio watermarks. Our robust watermarking scheme can recover the watermarks despite the packets loss (loss rate equal or less than 10%) on the networks. In addition, we integrate the classical Forward Error Correction (FEC) code in our watermarking scheme, which achieves 100% recovery rate when the packet loss is 6%.