Blind Audio Watermarking Scheme Based on Improved Cepstral Statistical Mean Modulation

Abstract

Audio watermarking is one of the best solutions to prevent the pirate audio propagation. The human auditory system is more sensitive than the visual system, which makes the audio watermarking more challenging. Many approaches for audio information hiding have been proposed. This paper designs a robust blind audio watermarking scheme based on modifying coefficients in cepstrum domain. According to the perceptual characteristics of human acoustics, time domain energy is firstly computed to select the embedding frames. Then, the watermark information is embedded in the complex cepstrum domain through improved statistical mean modulation method. Arnold scrambling transform is applied to ensure the security of the watermarking scheme. The proposed scheme achieves blind watermarking, since only the index of selected frames and the modulation threshold value are needed as secret keys in extraction process. Experimental results show that the designed watermarking scheme is suitable for all kinds of audio signals, for example, speech, music, ringtone, etc. This proposed method has good security, imperceptibility and robustness for different kinds of audio signals.