Research On ECG Signal Classification Algorithm Based On Machine Learning And Software Implementation

As people living and working pressure increase gradually,the cardiovascular disease has become the important factors threatening human life and health.Electrocardiogram can effectively inflect the health of heart,and it's a normal and effective mean that detects cardiovascular diseases.Electrocardiogram has important value in clinic,and is widely used in diagnosis of cardiovascular disease.So it's important that study a set of effective recognition classification algorithm of electrocardiogram signals.The main research contents of this paper are as follows:Firstly,band-pass filter is used to denoise the data in MIT-BIH database,wave group detection is carried out on the pre-processed data,R wave peak detection based on binary spline wavelet is improved,and adaptive threshold detection extremum pair is added to reduce R wave miss detection and false detection.A slope threshold platform search method is proposed to detect the peak and starting point of QS wave.Based on the detection information of QRS wave,the slope threshold platform search method is used to detect the peak and starting point of P and T wave,which improves the accuracy of wave group detection.The twelve eigenvalues such as RR interval and QT interval were calculated based on the detection information of wave group.The penalty factor and kernel parameters of support vector machine are optimized by genetic algorithm.Five kinds of electrocardiogram signals are classified based on eigenvalues and their performance is evaluated.The results show that the recognition and classification performance of support vector machine optimized by genetic algorithm is better and the accuracy is improved.Secondly,five kinds of electrocardiogram signals are classified based on convolution neural network.One-dimensional convolution neural network and two-dimensional convolution neural network models are designed,and they are evaluated the classification performance.The results show that the performance of convolution neural network is better than that of genetic algorithm support vector machine.The classification performance of two-dimensional convolution neural network is better than that ofone-dimensional convolution neural network,and the classification accuracy is improved.In this paper,we design a convolution neural network model to recognize and classify five types of electrocardiogram signals and six forms of clinical data collected from hospitals.The classification performance is good.The algorithm has good generalization ability and is suitable for actual data classification.In this paper,eight leads are identified and classified separately.The results show that lead II data has the best performance and can reflect electrocardiogram information more comprehensively.When it is impossible to obtain full lead data,lead II data can be used as reference.Finally,the electrocardiogram signal automatic analysis software is designed.Visual Studio 2013 is selected as the compiler and C # is used to program.The functions of data reception,data unpacking,interface display,recognition and classification and data storage are realized.The software is tested with data from MIT-BIH database.The test results show that the software can automatically recognize and classify electrocardiogram signals,and display the types and locations of anomalies.