A Fragile Watermarking Algorithm Based on Audio Content and Its Moving Average

Abstract

A fragile audio watermarking scheme for content authentication was proposed in the paper. First, two proper positive integers are chosen. Then, the audio signal is segmented. Two moving averages are computed according to the chosen integers in each segment. The watermark information obtained from audio segment is embedded at crosses of the two moving averages. The watermark can be blindly extracted without original signal when authentication is needed. Subjective and object tests show that the fragile watermarking scheme is inaudible and sensitive\e against MP3 compression, additive noise, re-sampling, re-quantization, low-pass filtering, etc. Especially, it can detect and localize the tampered regions effectively.