Abstract: A method of digital audio watermarking based on Independent Component Analysis is proposed in this paper.When embedding watermarking, the original audio is decomposed by Single-level discrete 1-Dwavelet transform, watermarking is embedded by mixing randomly to the approximation coefficient vector obtained, and the audio is Single-level reconstructed receiving audio having watermarking.When extracting watermarking, the audio is decomposed by Single-level discrete 1-Dwavelet transform, and watermarking is extracted by ICA based on kurtosis maximization.Emulated experiment results show that this method is robust and feasible.This method can protect the copyright of MRBT effectively on internet.