DCT-domain Watermark Detection Using Gaussian Mixture Model
基于高斯混合模型的DCT域水印检测方法

An audio watermark detector is developed based on the statistical characteristics of audio in DCT domain. Watermarks are embedded in DCT domain on a spread spectrum basis. A Gaussian mixture model (GMM) is introduced to model both the watermarked and unwatermarked audio signals in DCT domain. At the receiver a maximum likelihood detector is employed. Simulation results indicate that the proposed method allows for common signal processing and provides a better detecting accuracy compared with the linear correlation detector.