Abstract: Based on the psychoacoustic model Ⅱ and perceptual entropy (PE) used in MP3 encoding, watermark embedding is carried out by modifying the frames with PE above 1 800.By adding a random complex sequence to the spectral components in the medium and low bands in the frequency domain, a watermark is embedded.Extraction of the embedded data is achieved by using a frequency domain correlation approach without the original host signal.Experimental results show that the proposed scheme can achieve high imperceptibility and satisfactory robustness against MP3 encoding.

Key words: audio watermarking, psychoacoustics auditory model, MP3 encoding, PE