An algorithm of echo steganalysis based on power cepstrum and pattern classification

Audio steganalysis has attracted more attentions recently. Echo steganalysis is one of the most challenging research fields. In this paper, an effective steganalysis method based on statistical moments of peak frequency is proposed. Combined with power cepstrum, it statistically analyzes the peak frequency using short window extracting, and then calculates the eight high order center moments of peak frequency as feature vector. The SVM classifier is utilized in classification. All of the 1200 audio signals are trained and tested in out extensive experiment work. With randomly selected 600 audio signals for training and remaining 600 audio signals for testing, and with various embedding parameters combinations such as hiding segment length, attenuation coefficient, echo delay for hiding, the proposed steganalysis algorithm can steadily achieve a correct classification rate of 85%. Experimental results and theoretical verification show that this method is an effective method of audio echo steganalysis.

[1]  Ahmed H. Tewfik,et al.  Digital watermarks for audio signals , 1996, 1996 8th European Signal Processing Conference (EUSIPCO 1996).

[2]  Jinwoo Hong,et al.  New echo embedding technique for robust and imperceptible audio watermarking , 2001, 2001 IEEE International Conference on Acoustics, Speech, and Signal Processing. Proceedings (Cat. No.01CH37221).

[3]  Yôiti Suzuki,et al.  Time-spread echo method for digital audio watermarking , 2005, IEEE Transactions on Multimedia.

[4]  Walter Bender,et al.  Echo Hiding , 1996, Information Hiding.

[5]  Walter Bender,et al.  Techniques for Data Hiding , 1996, IBM Syst. J..

[6]  Hyoung Joong Kim,et al.  A novel echo-hiding scheme with backward and forward kernels , 2003, IEEE Trans. Circuits Syst. Video Technol..

[7]  Bin Yan,et al.  Improved echo hiding using power cepstrum and simulated annealing based synchronization technique , 2003, Proceedings of the 2003 International Conference on Machine Learning and Cybernetics (IEEE Cat. No.03EX693).