Application Research Of ECG Identification In Large Population

With the development of science and technology,people are increasingly demanding the security of personal information.Compared with traditional biometrics such as fingerprints,faces,and DNA,the mechanism of generating electrocardiogram signals is complicated,difficult to be copied,and high in safety.In the past research,the data commonly used in ECG signal identification research is the ECG signal from dozens to hundreds of people.Although the recognition accuracy is close to 100%,it is not universal.Applying the results to the actual situation is still lack of persuasiveness.The traditional identification method is applied to largescale crowds and it is not known what kind of problems will occur.This paper explores the effects and feasibility of applying ECG identification to large-scale populations,and the problems and solutions.In order to explore the problems that can occur when traditional methods are applied to large populations.The data set used in the experiment was an ECG record of 8,528 individuals from the PhysioNet 2017 Challenge,first using a waveform-based identification method,using four traditional similarity measures to identify the data using 50,100,and 150 in the PhysioNet dataset.The individual's ECG record,the recognition accuracy of the experiment is preferably 80%,37%,1.33%,it can be found that with the increase of the amount of data,the accuracy has dropped significantly.Aiming at the problem that the accuracy of traditional identification method decreases when the amount of data increases,this paper proposes a feature extraction method based on Matthew effect of ECG signal superposition matrix,which uses matrix to count the number of occurrences of ECG signals in the same area,effectively depicting the stable region of the ECG signal distribution to extract the stable characteristics of the ECG signal.In the experiment,the correlation coefficient was used to identify 7836 individuals,and the recognition accuracy was 88.72%.The recognition time of a single individual was 190 seconds.In the experiment using the feature extraction method and the correlation coefficient identification method of the ECG signal superposition matrix based on the Matthew effect,although a good recognition rate is obtained,the recognition time of a single individual is too long,and for the problem of recognition efficiency,This paper cites a local sensitive hash algorithm based identification method.The local sensitive hash algorithm uses the idea of hash collision to map the original space data to Hamming space,and uses Haiming coding to assign all individual ECG features.Different hash buckets.At the same time,the local sensitive hash algorithm utilizes the strategy of divide and conquer.When performing the recognition experiment,the individual to be identified only needs to perform matching calculation with other individuals in the same hash bucket without matching calculation with all individuals,thereby greatly improving the calculation.The recognition efficiency effectively reduced the recognition time of a single individual to 20.9 seconds,and the recognition accuracy rate was 92.37%.