Research On Encrypted Perceptual Hash Speech Retrieval Scheme Based On Feature Transformation

With the continuous development of multimedia information technology,the transmission of multimedia information has become an indispensable part of daily life.However,due to the large amount of multimedia information in the Internet space,how to accurately retrieve the required information from the massive amount of irrelevant information has become a hot issue in the field of multimedia information.This thesis cuts in from the speech retrieval technology,and conducts a series of researches on how to retrieve the required speech by using perceptual hash in the encrypted speech.This thesis mainly studies the use of feature transformation to generate perceptual hash sequence,speech encryption technology,speech retrieval technology and so on.The main work of this thesis is summarized as follows:(1)Aiming at the problem of low discriminability and poor robustness of the perceptual hash sequence generated by the existing feature extraction schemes,which lead to low efficiency of encrypted speech retrieval,a 2D-Gabor encrypted speech algorithm is proposed.The algorithm first pretreatment the speech,and extracts the characteristics of the speech through a 2D-Gabor filter to construct a perceptual hash sequence.At the same time,the speech is encrypted by the four-dimensional hyperchaotic system and the hybrid chaotic shift transform.Finally,the features of the speech to be retrieved are extracted and the perceptual hash sequence is generated,and the Hamming distance is used for matching retrieval.Experiments show that the hash sequence constructed by this algorithm has better discrimination and robustness,and can effectively improve the retrieval accuracy.(2)In order to meet the requirements of the security and accuracy of speech retrieval in the cloud environment and the discrimination and robustness of perceptual hash sequence,this thesis proposes a encrypted speech retrieval algorithm based on discrete K-L transformation.The algorithm first pretreatment the speech,and extracts the most representative characteristics of the speech through discrete K-L transformation to construct a perceptual hash sequence.At the same time,speech is encrypted by piecewise linear chaotic mapping and hybrid chaotic shift transform.Finally,the perceptual hash sequence generated by the speech to be retrieved is compared with the speech hash sequence library in the cloud.Experiments show that the hash sequence constructed by this algorithm has better discrimination and robustness,and improves the accuracy of encrypted speech retrieval.