Research On Identification Based On Pulse Wave Signal

With the development of society,the demand scenarios for identity recognition are becoming more and more diversified.How to find the most appropriate identification method in specific demand scenarios is the direction that people continue to explore in the field of identification technology.Biometric recognition technology,which uses the physiological characteristics and attributes of human body to identify,has gradually attracted people’s attention.Compared with the limitations of using external features for identification that are easily interfered by the environment and imitated,pulse signal,as an internal feature,is not easily affected by the external environment,and the acquisition device is cheap and easy to operate.This paper studies the identification method based on human pulse wave signal.The main research contents of this paper are as follows:Firstly,the generation mechanism of pulse signal and the working principle and parameter setting of pulse wave signal acquisition device are introduced.Combined with the formation principle of human pulse,the characteristics and physiological significance of pulse wave are analyzed and described.The interference sources and noise characteristics in the process of pulse signal acquisition are analyzed to facilitate the preprocessing of pulse wave signals in the next step.The existence of noise will affect the feature extraction of the pulse signal,and even cover up the important features contained in the pulse signal.In order to better analyze and process the pulse signal,it is necessary to preprocess the pulse wave signal.In this paper,the principle of wavelet transform and wavelet multi-resolution property are used to denoise the pulse wave signal.In the process of noise reduction,the decomposition levels are set based on the frequency characteristics of the noisy pulse signal,and the effects of different wavelet basis functions and threshold functions on the noise reduction of the pulse signal are compared.The filtered pulse signal is divided into cycles,and the abnormal signals in the pulse wave signal are filtered and eliminated based on the difference value hash algorithm.Signal feature is the concentrated expression of signal information,and the accuracy of feature extraction directly affects the effect of signal recognition.In this paper,the time-domain characteristics of pulse signal are extracted based on the physiological meaning of pulse signal waveform.In view of the adaptability of Hilbert transform in time-frequency analysis,the pulse signal is decomposed into multiple natural mode components by using the key steps of Hilbert transform to set empirical mode decomposition,and the time-frequency domain characteristics of the pulse signal are obtained by extracting the statistical features of the natural mode components.BP neural network is used to recognize the extracted feature vector.The pulse wave signal is identified based on the convolutional neural network.Considering the feature extraction ability of the convolutional pooling operation of the convolutional neural network on the input signal,the signal is no longer extracted separately in the early experiment.By constructing a parallel multi-channel onedimensional convolution network model,the filtered pulse signal and the intrinsic mode components obtained by the collective empirical mode decomposition are taken as the input of the model,and the feature extraction ability of the convolution layer and pooling layer is used to extract the features of the input data,and then the full connected network is used to identify.On the basis of the network model structure,the attention mechanism module is added to further improve the recognition efficiency of the model by giving different weights to different channels.