In order to study the audio digital watermark robustness and perceptuality

the particle swarms opti- mization algorithm is introduced into the spread-transform dither modulation algorithm. Through spread-transform dither modulation algorithm achieve audio digital watermark can be embedded and blind extracted. Three levels of discrete wavelet transform were carried out on the carrier audio

and discrete cosine transform on low frequency component

in the hybrid domain watermark embedding

and according to embed information adaptive selection em- bedding location. By particle swarms optimization algorithm to spread-transform dither modulation algorithm quanti- zation step length and the number of projection elements for the optimal choice. With the combination of the signal- noise ratio and normalized correlation we construct the fitness function

balance the contradiction between the per- ceptual and robustness. The experimental results show that the algorithm has a good imperceptibility

and robust- ness.