Long Sequence Speech Perception Hash Authentication Based on Multi-feature Fusion and Arnold Transformation

Yi-Bo Huang ； He-Xiang Hou ； Man-Hong Fan ； Wei-Zhao Zhang ； Qiu-Yu Zhang

《International Journal of Network Security》 23卷2期 (2021/03) Pp. 359-370

https://doi.org/10.6633/IJNS.202103_23(2).20

摘要

To improve the discrimination and robustness of the existing speech authentication and solve low security in the process of mobile communication transmission, a novel long sequence speech perceptual hash authentication algorithm based on multi-feature fusion and Arnold transform is proposed. Firstly, the wavelet low-frequency logarithmic energy spectra of the pre-processed speech signals and the feature matrix of the low-frequency MFCC are extracted. Secondly, the two sets of feature matrices are transformed into binary hash long sequences. Finally, the two long hash sequences are fused into a novel long hash sequence after Arnold encryption to complete the hash matching. The proposed algorithm adopts a hash long sequence, which significantly improves the discrimination of existing algorithms. When each frame of the speech signal is converted into a binary hash sequence of 8 bits, the algorithm's robustness is virtually balanced. Experimental results show that compared with the existing speech authentication algorithms, the proposed algorithm has better comprehensive performance and ensures the security of the hash sequence in the transmission process.

關鍵字

Arnold Transform ； Discrimination ； Perceptual Hash Long Sequence ； Speech Authentication ； Wavelet Logarithmic Energy Spectrum