Perceptual audio data concealment and watermarking scheme using direct frequency domain substitution

A method is described for perceptually transparent data concealment and watermarking in audio. The proposed system replaces redundant and imperceptible frequency components with hidden data. A psychoacoustic model is used to identify suitable frequency locations for data hiding. Such a method is complicated by the windowing and overlapping requirements used for signal conditioning. The proposed system uses data flipping in place of windowing and incorporates a novel data detection scheme with adaptive weighting to increase the robustness of the watermark transmission. The resistance of the watermarking system to filtering, amplitude scaling and additive white noise is measured and results presented.