According to the various formats cause the phenomenon of audio file's watermarking embedding becomes complex and complicated,this paper used FFMPEG to make the various formats transform into a unity format to embed the watermarking which improved the efficiency and result. it selected improved quantizing discrete cosine transform algorithm to embed the watermarking. First,it converted the binary image into the one dimensional sequence and modulated the sequence with scrambling encryption and polarization operation,then performed the digital audio signal the quantizing DCT transform,according to each different subparagraphs energy grade,it used different embedding strength and selected four different levels coefficients, repeated embedding. In the experimental simulation results,there doesn't exist obvious differences between the transformed audio signal and the original audio signal,the watermark algorithm is imperceptible to digital audio signal such as Gauss white noise interfere,low pass filter and loss compression.