Biometric Templates In Speech Authentication And Retrieval Algorithms

With the improvement of the performance of mobile phones,computers and other portable devices and the rapid development of the Internet,today's speech information increases exponentially,and there are many kinds of speech formats.This change makes it urgent for people to realize speech authentication in multi-format,improve the efficiency and accuracy of speech retrieval and ensure the security of cloud data.Based on the background of multi-format speech authentication and encrypted speech retrieval,this paper studies digital signal processing,biometric template construction,speech encryption and hash index construction algorithm.The specific research is as follows:1.In view of the problems that the application scope of existing speech content authentication is small,and the robustness,discrimination,security and real-time authentication need to be improved in the authentication process,this paper proposes three multi-format speech authentication algorithms based on perceptual hash and two multi-format speech authentication algorithms based on Bio Hashing through the research of biometric template and encryption algorithm.The algorithm firstly extracts the speech clip,then uses the perceptual hash method and the Bio Hashing method to extract the feature matrix of the biometric template to generate the hash index,and finally uses the encryption algorithm to encrypt the hash index with equal length.Experimental results show that the algorithm has the characteristics of good security,high discrimination,realtime and wide application range,and the comparison method verifies that the Bio Hashing has the unidirectional property of trapdoor,and realizes the detection and localization of speech tamper in a small range by the minimum code distance method.2.Aiming at the leakage of plaintext data in the existing speech Bio Hashing retrieval system,and improving the accuracy of speech retrieval,the diversity and revocability of biometric templates,a long-sequence Bio Hashing encrypted speech retrieval based on feature fusion is proposed by studying the revocable biometric template and speech encryption algorithm.The client first establishes a biometric template with a single mapping key according to the classification of speech features.Then,the chaos measurement matrix generated by the feature vector and key is used to generate the feature index of speech by using the Bio Hashing method,and the improved SHA256 algorithm is used to encrypt the speech file.Finally,the feature index and encrypted speech are sent to the cloud.Experimental results show that the algorithm can not only effectively prevent the leakage of plaintext,but also has good diversity and revocability of biometric templates.At the same time,the algorithm not only has good efficiency and accuracy,but also solves the problem of speech retrieval after content preservation operation.