Speech Authentication Based on Perceptual Fingerprint Hashing

A perceptual fingerprint hashing method is proposed for audio authentication in this paper. First, the biologic fingerprint image is divided into overlapping rectangle blocks by randomly using a chaotic sequence, and then the gravity center of each block is calculated. Finally, the perceptual Hashing value is generated based on the gravity center of each block. To authenticate the speech identity source, the perceptual fingerprint hashing that represents speaker’s identity information and speech signal must be connected closely. So the perceptual fingerprint hashing value is embedded into the speech signal as a watermark. At the authentication stage, first, the perceptual fingerprint hashing value is extracted from the speech for authentication, and then it matches another perceptual fingerprint hashing value which is restructured by the suspicious fingerprint image. Finally, the authentication result can be decided by the matching result. The experimental results indicate that the proposed algorithm is not only robust against fingerprint rotation processing and noise attack, but also has high security and uniqueness.