In this paper, we present a novel self-synchronization algorithm for audio watermarking to provide copyright protection against different kinds of audio attacks, especially, the synchronization attacks. A special region centered at the feature point of the original audio signal was exploited to embed the watermark information. The feature point we chose can survive even after most kinds of attacks, thus the self-synchronization ability can be obtained by using the feature point as synchronization signal. To embed the watermarks into the low frequency coefficients in DWT (Discrete Wavelet Transform) domain, high robustness and inaudibility could be guaranteed. A watermarking technique in the category of quantization index modulation was adopted in this algorithm due to its great robustness and blind feature. The performance of this proposed algorithm was evaluated by PSNR (Peak Signal Noise Ratio) and Normalized Cross Correlation (NC) after various attacks. Simulation results show that this algorithm is robust against not only synchronization attacks but also various common attacks such as filtering, resample and so on. © 2012 International Information Institute.