General Regression Neural Network Based Audio Watermarking Algorithm Using Torus Automorphism

Accurate extraction of embedded data at the receiver end is still a major point of consideration in audio watermarking area. This paper portrays a blind audio watermarking scheme in transform domain using the combination of properties of audio signal extracted through singular value decomposition and general regression neural network leading to exact extraction of watermark. The security of embedded watermark is assured by using torus automorphism at the embedded side. Results from the experimental setup validate the accuracy of proposed scheme. The payload capacity of proposed algorithm is 62.5 bps. The comparison of proposed scheme with existing ones indicate that the proposed scheme has shown good efficiency in terms of robustness, payload and transparency.