ECG Identification Based On Gradient Enhancement Machine Learning Algorithm

Nowadays,technology has changed our way of life,especially strict requirements for information security.New identification techniques are needed to improve information reliability.The new biometrics technology is safe and convenient,provides great convenience for people,and has been widely used in the market.Ecg signal is one of the biological signals.It has been used in the medical clinical field before.Due to its unique advantages,it is gradually applied in the identification technology.The unique advantages of ECG features mainly include the following two points: first,ECG signal is an in vivo feature,which is difficult to steal features,and the ECG acquisition technology has a long history and is mature;second,ECG signal is a one-dimensional signal,which is easier to process than other biological features.At present,the research on ECG identification has made a lot of progress,but there are still some shortcomings such as feature redundancy,low accuracy and poor timeliness.We conducted the following research on the above issues:1.An ECG identification scheme based on ICA-CatBoost was constructed.Gradient Boosting(GB)shows great advantages in prediction and classification due to its good compatibility,and provides technical support for establishing a safe and reliable ECG identification model.In view of the existing problems of feature redundancy and low accuracy at home and abroad,an ECG identity recognition algorithm based on ica-catboost is constructed in this paper,which is a machine learning framework based on gradient elevation decision tree.The original ECG signal contains noise,andsignal-noise ratio are relatively low.Firstly,ECG signal was filtered by a filter or wavelet transform,and then the ECG signal was segmented by differential positioning at the legal R point to obtain single-period ECG.After pretreatment,ECG signals were processed by Independent Component Correlation Algorithm(ICA)to extract the effective information of the high-order components by which multi-dimensional data were converted into low-dimensional data.Finally,CatBoost classifier was used for recognition.This scheme effectively solves the redundancy of ECG signals and combines with CatBoost classifier to greatly improve the accuracy and practicability of identification.The experiment shows that theCatBoost algorithm can greatly improve the accuracy of the system.Compared with Support Vector Machine(SVM),KNN,Linear Discriminant Analysis(LDA)and random forest,the proposed method achieves higher accuracy which is 98.05%.2.An ECG identification scheme based on Light GBM algorithm was constructed.Aiming at the time consuming problem of hot coding for all features in the above system,the efficiency problem of scanning all sample points in order to find the optimal segmentation point in large samples and high latitude data was solved.Light GBM algorithm was introduced to build the classification model.The Light GBM algorithm uses the method of calculating gradient and binding mutually exclusive features of sample sampling to avoid using all samples.The independent component analysis algorithm is used to reduce the dimensionality and then imported into the Light GBM algorithm model.The objective function of this algorithm model adopts histogram differential acceleration method,which has faster training efficiency and low memory usage,and is more suitable for industrial practice.The proposed method is based on ECG data validation in ECG-ID and PTB database.Experiments show that the Light GBM algorithm improves the recognition accuracy and speed,and the recognition rate is 99.02%,which makes the algorithm more suitable for practical application.Although the application of neural network algorithm has been widely promoted in recent years,the improvement of the algorithm is more obvious in the case of limited training samples,short training time and lack of relevant parameter adjustment.The research of this thesis has some advantages in improving the accuracy and speed of ECG identification model,and makes ECG identification more practical.