关键词: 音频水印，MDCT频谱嫡，隐含同步，特征识别

Abstract: An innovative robust audio watermarking scheme based on using MDCT spectral entropy as implicit synchronization to achieve self-synchronization was proposed in this paper. The algorithm makes full use of MP3 compression mechanism to implement reduplicate embedding on audio segments after decoding, combining the masking effect in psychoacoustic model, determines the suitable low-frequency DWT coefficients to deal with. Signals are recompressed to form MP3 bitstrcam which contains watermarking information. Extracting can accomplished without the original audio and is guaranteed by identifying the sequence of MDCh spectral entropy. hhe experimental results show the proposed scheme itself can achieve good excellent robustness to desynchronizaition attacks and common audio processing methods created by Stirmark.

Key words: Audio watermarking, MDCT spectral entropy, Implicit synchronization, Feature recognition