Audio watermarking

Abstract

This project proposes a blind digital audio watermarking scheme for the purpose of protection of intellectual property. The watermark is embedded in the Discrete Cosine Transform (DCT) domain. The psychoacoustic model is incorporated to ensure the imperceptibility of the embedded watermark. The watermark is recovered without any reference to the original audio signal. A novel synchronization scheme, the peak point extraction (PPE) scheme, is proposed in this project to be the prerequisite of the blind recovery. The idea of error correcting codes is borrowed from digital communication to improve the rate of correct watermark recovered. A linear block code, one type of the error correcting codes, is implemented in this project. The proposed digital audio watermarking scheme possesses excellent imperceptibility of embedded watermarks. The proposed scheme also shows good robustness against attacks such as cropping and resampling.