It is a challenging work to design a robust audio watermarking scheme against desynchronization attacks.In this paper，a feature-based digital audio watermarking scheme robust against desynchronization attacks is proposed.Firstly，the steady feature points are extracted from the host audio according to the local energy distribution.Then，the candidate audio segments for embedding watermark are defined by using the feature points.Finally，the digital watermark is embedded into a host audio by modulating the statistics average value of audio samples.In digital watermark detection，the feature points are selected by the same technique as the embedding，and the digital watermark is extracted from the watermarked audio after the feature points.Experimental results show that the proposed scheme is inaudible and robust against common signals processing such as MP3 compression，low-pass filtering，noise addition，and equalization etc，and is robust against desynchronization attacks such as random cropping，amplitude variation，time-scale modification，and jittering etc.