An audio watermarking of wavelet domain based on BP neural network

This paper proposes an audio watermarking algorithm of wavelet domain based on BP neural network We introduce BP neural network into audio watermarking systems, embed watermark in the low-frequency, important components of wavelet coefficients. In order to improve the efficiency and accuracy of BP network training, design a high-efficient, high- precision, fast training algorithm of BP neural network, which improve the performance of the watermarking system effectively. Experimental results show that the capacity of the proposed scheme is relatively high-transparent and high-robust the improved fast training BP algorithm converges fast, reaches high accuracy.