Research On Identity Recognition Method Based On Pulse Wave Signal

The development of network technology brings convenience and efficiency as well as many challenges.Various network systems have penetrated into all aspects of our lives.These systems require risk control,and identity verification is a key means to implement access control strategies.Identity verification is ubiquitous in daily life.In the case of many and complex applications,ensuring the safety,efficiency and accuracy of identity recognition has become a difficult problem in the field of identity recognition.In this context,the identification technology based on biological characteristics(such as face,fingerprint,iris,voice,etc.)has been greatly developed.Biometrics technology solves the problems of lost,forgotten,and leaked user passwords.It is generally difficult to forge and has high security.However,each type of biometric technology has its own applicable scenarios,and the existing biometric technology has its limitations in special scenarios such as open construction environments.This kind of scene requires non-sensitivity and normality verification.The pulse wave signal,as a biological signal of the living body,cannot be copied and has high security.And the collection is simple and portable,and there are many mature commercial products.Using the pulse wave signal as a feature for identity recognition can effectively ensure the accuracy of recognition and meet the portability and high degree of freedom requirements of the above-mentioned special scenes.From a medical point of view,introduce the reasons for the formation of pulse waves,the basic characteristics of the waveform and the principles of acquisition,and demonstrate the use of pulse wave signals for identification through aspects such as the current research status of pulse waves for identification and the characteristics of biometrics that can be used for identification.The feasibility.Analyze the characteristics of the pulse wave signal and the source of noise,design a noise reduction plan,use band-pass filtering,morphological filtering,wavelet multiresolution analysis filtering and other methods to reduce the noise of the collected data,and compare and analyze the processing results.Finally,the wavelet multi-resolution filtering method is selected as the noise reduction method of pulse wave signal data.Put forward the idea of increasing the dimension of pulse wave signal data,introduce five methods of short-time Fourier transform,recursive graph,Markov transition field,Grammy angle and field to increase the dimension of the pulse wave data,and write the pulse wave signal into one dimension.The structure is converted into a data structure in the form of a two-dimensional image that is highly correlated with time and can reflect the changes in pulse wave details over time,that is,a grayscale image that only contains information about the amplitude and angle of time.The data set is made with the upscaled data,and the residual neural network that has excellent performance in the field of image classification is used,and the three aspects of Resnet18 adaptive learning rate,parameter initialization,and fully connected layer replacement are optimized.Design a pulse wave identity recognition scheme,and use the model to perform identity verification experiments on the testees.Experiments have shown that the Grammy angle difference field data has the highest recognition accuracy after ascending.Through the analysis of multi-period and single-period pulse wave identification experiments,because the multi-period pulse wave data contains more pulse wave characteristics,the recognition effect is better,and the final recognition accuracy rate is as high as 96.41%.