A novel watermarking algorithm for protecting audio aggregation based on ICA

This paper proposes an audio watermarking algorithm based on ICA, breaks through the traditional watermarking which only protects the copyright of a single audio, implements copyright protection for an audio aggregation. The implementation process of the algorithm is composed of following steps: firstly, the wavelet coefficients of each audio work in the aggregation are set as observation data, and independent components of the observation data are extracted using ICA separate matrix, and then two most important components are chosen to embed the common watermark; secondly, ICA is used to detect watermark. According to experimental results, after embedding watermark into original audio aggregation, there is little impact on audio aggregation´s perceptual quality, and when detected audio aggregation suffered from various attacks, the algorithm can extract watermark from the attacked detected audio aggregation. The proposed algorithm has good imperceptibility, strong robustness, low computational complexity, and achieve blind detection.