The Research Of Speech Signal Encryption Based On Compressed Sensing

With the rapid development of science and technology in recent years,the society has entered the era of big data.The communication becomes more and more frequent and the amount of data is becoming more and more.As a unique human signal,speech signal can carry more information.So people pay more attention to the security of speech information.How to encrypt the speech signal after digitizing the speech signal has always been a hot topic.Compressed Sensing(CS)as a new sample theory in recent years can take advantage of the compressibility of speech signal to process and analysis speech signal.So research the high efficient and security encryption method of the speech signal encryption based on Compressed Sensing has high theoretical and practical sign ificance.In this paper the main work is as follows:The basic mathematical model of Compressed Sensing is introduced and the Compressed Sensing is divided into three parts.The reasons for the sparse representation of signals and some common sparse representation methods are introduced.This paper introduces the reasons for the observation of the signal and how to determine the feasibility of the observation matrix and also introduces some common observation matrix.This paper introduces some common reconstruction algorithms and the characteristics of each algorithm.Then I also introduce the problem of compression perception which needs to be studied in the future.Standard Compressed Sensing speech encryption model is put forward.This paper introduces the production and characteristics of speech signal and how to separate the surd and dullness signal by the characteristic parameters.After the pretreatment,according to the standard encryption model analyzes the sparse performance of the surd and dullness signal of the male and female voices in discrete cosine transform,discrete Fourier transform and other orthogonal transform to verify speech signal can be encrypted.I research observation performance of the speech signal in the random gaussian matrix,random Bernoulli matrix and other common observation matrix and select the appropriate observation matrix as the key matrix to encrypt the speech signal.This paper analyzes the influence of the frame length and the key matrix's compression ratio on the signal reconstruction.Analyzes the various greedy algorithms and selects the appropriate algorithm and related parameters for the standard encryption model.I put forward a model of speech chaotic encryption based on Compression Sensing.According to the analysis results of standard Compressed Sensing speech encryption model I choose the appropriate key matrix and related speech parameters.Initial sensitivity of chaotic system is used to increase security.I use the cat mapping to complete the scrambling of information,then complete the encryption process.The decryption process will reverse the process.Experimental simulation of this model was conducted under random signal and the male and female speech signal.Then I analysis perfect confidentiality and actual confidentiality of this model.The test improve that the model has good actual confidentiality.