Research On Efficient Retrieval Method For Encrypted Speech Based On Perceptual Hashing

With the increasing popularity of multimedia collection devices and the rapid development of data storage as well as Internet technologies,multimedia data has become the main medium for people to acquire and disseminate information.How to ensure the security of users' data and how to protect data in the face of massive multimedia data under the conditions of accurate and rapid retrieval of the required content from the massive data,has been a hot issue in the field of multimedia retrieval research.Based on the content based encrypted speech retrieval technology as the starting point of the study,the speech encryption method,efficient speech feature extraction,speech classification,retrieval index structure,retrieval algorithm and other key technologies were studied.The main research contents are as follows:1.Aiming at the problems of low efficiency of speech feature extraction of existing methods,poor robustness and discrimination of constructed perceptual hashing scheme,and low retrieval accuracy,an efficient encrypted speech retrieval algorithm based on frequency band variance is proposed.The algorithm firstly pre-processes the speech,calculates the frequency band variance of the speech as the speech feature to construct the hash sequence;then encrypts the speech by Logistic chaos scrambling;finally,extracts the hash sequence from the query speech,and uses Hamming distance algorithm to perform matching retrieval.The experimental results show that the hash sequence constructed by the proposed algorithm has better discrimination and robustness,and effectively improves the retrieval performance.2.Aiming at the defects of the traditional traversal matching retrieval algorithm and the realization of massive encrypted speech retrieval,an efficient encrypted speech retrieval algorithm based on IFFT transform and measurement matrix is proposed.Firstly,the speech perceptual hashing sequence is constructed by combining the IFFT transform and measurement matrix,and the Hennon chaotic scrambling encryption is performed on the speech.The speech is then classified and the generated hash sequence is compressed by the run length compression technique and uploaded to the cloud system hash index table.At the time of retrieval,the matching retrieval is performed by the Hamming distance algorithm.The experimental results show that the robustness,discrimination and feature extraction efficiency of the proposed perceptual hashing scheme are better than the existing ones.It has a good recall and precision ratio,and with high retrieval efficiency andaccuracy.3.In order to sufficient the requirements of efficient and secure retrieval of encrypted speech data in cloud environment,and the robustness and discrimination requirements of perceptual hashing scheme,an encrypted speech retrieval algorithm based on Chirp-Z transform and perceptual hashing second feature extraction are proposed.The algorithm first performs Chirp-Z transform on speech after pre-processing,filters it by pseudo-random sequence,constructs a hash sequence as a speech feature,and encrypts the original speech through m-sequence-based encryption;Then each speech perceptual hashing sequence is uniquely mapped to a decimal number through quadratic feature extraction,and uniformly classifies the speech by k-means clustering to construct a system hash index table.Finally,the query speech is processed by denoising and matched by the Hamming distance algorithm.The experimental results show that the proposed algorithm greatly compresses the information capacity of speech features,significantly improves the retrieval efficiency,and has a good retrieval effect on noisy speech.