TAO Zhi¹  ZHAO Heming²  GU Jihua¹  WU Di²

(1  School of Phys. Sci. and Tech., Suzhou Univ.  Suzhou  215006)

(2  Dept. of Electron, School of Information, Suzhou Univ.  Suzhou  215006)

Received Jul. 15, 2005

Revised Jan. 23, 2006

Abstract  Watermark embedding algorithm based on critical band wavelet transform of digital audio signal is proposed in this paper. The masking threshold for each audio signal segment was calculated on the basic of psychoacoustic model. According to the similarity between critical band of human auditory system and critical band wavelet transform, a watermark was embedded into the low-band and mid-band coefficients of digital wavelet. The embedding strength was adaptively controlled by the masking threshold. The experiment results show that the embedded watermark signal is inaudible, and the watermarked audio signal has good robustness against many attacks such as compression, noise, re-sampling, low-pass filtering.

PACS number: 43.60, 43.70