Self-Synchronous Time-Domain Audio Watermarking Based on Coded-Watermarks

A time-domain audio watermarking algorithm based on low-frequency amplitude modification was proposed by Lie in 2006. The algorithm has acceptable robustness against common attacks and it also achieves synchronization. However Lie´s method has two disadvantages: the coded watermark was produced by convolutional codes which are not the best option for very noisy channels like watermarking ones and also, the payload is decremented because the synchronization code. We propose an improvement to Lie´s algorithm which solves the problems abovementioned. Our improvement replaces both synchronization code and convolutional code for only a self-synchronous decoding algorithm which uses low-density parity-check (LDPC) codes. With this modification, higher payload is obtained because synchronization codes are avoided, and the watermark is more robust due to the LDPC code. The self-synchronous decoding is based on the good cyclically permutable (CP) characteristics of LDPC codes. Our improvement is capable to kept in sync with combined attacks, e.g. MP3 and clipping attack or low-pass filtering and clipping attack.