Research On Secure Hash In Low Complexity Speech Retrieval Algorithms

The rapid development of mobile Internet has led to the high dimension and complex space of speech data.Existing speech retrieval algorithms cannot meet the requirements of efficient retrieval performance and privacy security of speech data in massive applications.With the continuous growth of dimensions in retrieval system,only a few data points are distributed near the center(empty space phenomenon),and the distance between data points in high-dimensional space is nearly equal(dimension effect).This thesis focuses on the search performance and security issues caused by empty space phenomena and dimension effect.The specific research contents are as follows:1.Aiming at the problems of retrieval accuracy,efficiency and the security of speech storage and transmission process,this thesis proposes a security framework based on KNN secure hash for speech retrieval algorithm.This algorithm takes the spectral centroid of speech as the only input factor,and then uses KNN classification to train and learn the speech vector and obtain each speech centroid,and assigns a specific hyperchaos Lorenz-compression sensing encryption algorithm key to each speech centroid,builds a security framework based on the cancellable biometric template generated by combining classification and distribution key.The experimental results show that the algorithm improves the retrieval efficiency and accuracy.The operation of generating biometric template for hash vector and classification increases the diversity,cancellability and security of the template.The speech encryption effectively prevents the disclosure of clear text,ensures the security of speech storage and transmission process.2.Aiming at the problems of retrieval computation complexity,content verifiable retrieval after speech attack,empty space phenomenon and dimension effect,this thesis proposes a low complexity speech secure hash retrieval algorithm based on KDTree nearest neighbor search.The algorithm extracts the spectrum subband centroid of speech to generate the feature vector,and then uses KDTree classification to establish the index of the cancellable biometric template,and assigns a specific SHA256-Ushiki chaotic encryption algorithm key to each index.According to the cancellable biometric template generated by the combination of classification and distribution key,the security framework is constructed.The experimental results show that through the establishment of the KDTree cancellable biometric template index,which effectively solves the empty space phenomenon and dimension effect.Through the KDTree nearest neighbor search,the algorithm reduces the number of matching between classes,and only needs to match once within class,effectively reducing the computational complexity of the retrieval.Compared with the tampering of mobile terminal,the content of speech can be verified and retrieved.On the basis of traditional speech retrieval algorithms,this thesis proposes two low complexity speech secure hash retrieval algorithms with the goal of building efficient,accurate,secure,and robust speech retrieval,and solves the computation complexity and security issues caused by dimension effect and empty space phenomenon.In the following research,we will optimize indexing methods to improve retrieval performance.