Research On Wearable Bioelectric Signal Identification System

With the development of science and technology,while information technology is convenient for people's lives,it also brings about personal information security issues.The existing biometric identification has the disadvantages of being easy to be stolen,easy to forge or imitate,unstable.Compared with the existing biometrics,identification based on heart signals has more advantages such as concealment,safety,and stability,and has become one of the hotspots of current research.This thesis mainly studies and designs a wearable bioelectrical signal identification system,which takes the electrocardiographic(ECG)and photoplethysmography(PPG)in the heart signal as the research objects.In the study,a wearable remote continuous identity recognition system based on compressed sensing was designed to realize individual identity recognition based on ECG signals and PPG signals.The main research contents of this thesis include:(1)A wearable heart signal acquisition system is designed to achieve continuous acquisition of human heart electrical signals and pulse signals,while avoiding the constraints of the acquisition system on individual activities.The system circuit mainly includes: ECG electrode sensor,PPG signals sensor,analog-to-digital conversion circuit,STM32 control circuit,power supply circuit,bluetooth transmission circuit,etc.The hardware circuit structure of the system is simple,the measurement is accurate,and it is easy to carry.(2)The redundant dictionary constructed by K-SVD is compared with the discrete cosine transform,and the redundant dictionary is selected as the best sparse representation method.The advantages and disadvantages of different observation matrices in this system are analyzed.Based on the advantages of random characteristics,low power consumption and easy hardware implementation,a sparse binary random observation matrix is constructed.The reconstruction performance of ECG and PPG signals under the boundary sparse bayesian learning based on the bound-optimization-based block sparse bayesian learning(BSBL-BO)algorithm and stage weak orthogonal matching pursuit(SWOMP)algorithm are studied.(3)The noise source and waveform characteristics of ECG signal and PPG signal are analyzed,and the two signals are pre-processed by wavelet denoising method.Different feature extraction schemes are designed according to the different characteristics of the two signals,and stable and accurate signal characteristics are obtained.Using support vector machines to classify and identify individuals,experiments show that the identification based on the two heart signals has reached a high recognition rate,and verified the feasibility of the compression and reconstruction scheme used in this thesis.