New detection scheme for spread spectrum audio watermarking

Abstract

This paper presents a new improved scheme to the detection algorithm for the spread spectrum audio watermark. Based on the traditional detection method, our detection algorithm normalizes the correlation sequence and then compares its energy against a threshold to determine whether an audio segment is watermarked or not. Because this energy is independent of the amplitude of the audio segment, the detection is not affected by the audio signal and thus the detection error rate is very low. Experimental results show that the proposed detector makes great progress on the robustness to normal attacks.