This paper presents a reversible data hiding scheme for digital audio by using noncausal prediction of alterable\norders. Firstly, the samples in a host signal are divided into the cross and the dot sets. Then, each sample in a set is\nestimated by using the past P samples and the future Q samples as prediction context. The order P + Q and the\nprediction coefficients are computed by referring to the minimum error power method. With the proposed predictor,\nthe prediction errors can be efficiently reduced for different types of audio files. Comparing with the existing\nseveral state-of-the-art schemes, the proposed prediction model with expansion embedding technique\nintroduces less embedding distortion for the same embedding capacity. The experiments on the standard audio\nfiles verify the effectiveness of the proposed method.
Loading....