Efficient Robust Speech Authentication Algorithm for Perceptual Hashing Based on Hilbert-Huang Transform

Abstract

According to the resource-constrained of mobile communication terminal, the traditional broadband speech authentication algorithm cannot satisfy the authentication of a small amount of data, strong robustness and high e‐ciency, etc. This paper, using the short-time stable characteristics of speech signal, proposed an e‐cient robust speech authentication algorithm for perceptual hashing based on Hilbert-Huang Transform (HHT). Firstly, the algorithm obtains to the HHT coe‐cients of each speech frames and constitutes an HHT coe‐cients matrix after pre-processing, framing and adding window to speech signal. Then, the HHT coe‐cients matrix for two-dimensional Discrete Cosine Transform (DCT) generates DCT coe‐cients. It based on DCT coe‐cient characteristics to evaluate its entropy. Finally, a quantization step gets authentication perceptual hash sequence. Experimental results show that the algorithm, compared with the existing methods, has a good robustness and discrimination for content preserving operations, and the short authentication time. Also, it can meet the real-time requirement of the speech perceptual authentication.