A robust audio watermarking algorithm based on frequency hopping

In this paper a novel audio watermarking algorithm based on the idea of frequency hopping is presented. Two separate sets of different frequencies are used to transmit message bits `0' and `1'. These frequencies have been selected to tolerate different attacks while keeping the watermark signal inaudible. To embed each bit of the binary message a pair of frequency is selected randomly from the given sets for each audio segment. One frequency is picked up from the pair based on the message bit and the frequency content of the audio segment. A sinusoidal carrier is generated at this frequency and is used as the watermark in the segment. The amplitude of this sinusoidal carrier is shaped based on critical band filters which models the human auditory system (HAS). The content based properties of the proposed scheme offers maximum flexibility while satisfying robustness and high capacity, simultaneously. Simulation and experimental results show high resistance of the proposed algorithm against all known watermarking attacks for different kinds of audio clips.