Optimized audio watermarking scheme with swarm intelligence

This paper presents a digital audio watermarking scheme that can embed imperceptible and robust watermark image into the audio signal. The watermarking process is performed in discrete wavelet domain by applying the QR factorization on 2-D blocks of low frequency part. The watermark bits are inserted in each block by quantizing the coefficients of resulting R matrices. A particle swarm optimization technique is designed to find an optimal quantization step which allows to obtain the highest possible robustness without losing the transparency of the watermark. A series of experiments has been carried out by using different host audio signals under different attacks. Results show that the audio quality is maintained and the watermark remains robust despite various signal processing attacks. In comparison with some recent audio watermarking algorithms, our method presents competitive results in terms of imperceptibility and robustness.