ECG Feature Detection And Arrhythmia Classification

Heart disease is one of the main killers that endanger human life and health. It has the characteristics of strong concealment, rapid onset, short response time and high mortality. However, there are still susceptible to noise interference in the existing automatic analysis and detection technology, for the problems such as the complex morphology of arrhythmia classification is not accurate.In view of these problems, this paper solves the problem of noise preprocessing, ECG feature detection and classification model design. Aiming at the characteristics of the types of noise in ECG signals and the wide coverage of frequency domain, the original signal is filtered by nine layers of wavelet transform filter. The ECG characteristics of wave shape is complex and changeable, amplitude variation characteristics of proposed a fusion k nearest neighbor thought of adaptive threshold characteristics of ECG signal detection method, and the use of swarm intelligence particle swarm optimization(PSO) algorithm for parameter optimization. For the automatic classification of arrhythmia, according to the characteristics of MIT-BIH data set design based on decision tree classification SVM model, and the PSO algorithm to model the input feature selection and model parameters optimization and of normal sinus rhythm, left bundle branch block LBBBB, right bundle branch conduction block RBBBB, real premature beat(APC) and ventricular early shrinkage(PVC) five types of heart beats automatic classification.The nine layer decomposition wavelet transform filtering algorithm compared with the traditional band pass filter, can retain the details of the original wavefo rm information under the condition of filter of ECG baseline drift, EMG interference, frequency interference noise. K nearest neighbor adaptive threshold method compared to the existing algorithm with accuracy, strong real-time performance, application range wide characteristics, especially for affected by noise seriously, the poor quality of the acquisition of signal, can provide high accuracy of detection. For arrhythmia classification method, combining the PSO algorithm and support vector machine(SVM) classification algorithm, in order to ensure the classification accuracy under the condition of a reduction in the number of input features, than the traditional detection algorithm has stronger detection accuracy and generalization ability. These work for future field of ECG signal detection algorithm design and optimization, disease auto diagnosis of clinical application, has very important theoretical and practical significance.