Research Of Feature Extraction And Classification For Electrocardiogram Based On Spline Curve

Heart disease is still one of the most terrible threats to human health. The automatic diagnosis of electrocardiogram can effectively help doctors diagnose and treat various heart diseases. As a consequence, the research of ECG has been a hotspot among scholars all over the world.The research objectives in this thesis are patients’original ECG data collected from hospitals. A new method is presented in this thesis to extract features through three-order B-spline curves and to cluster ECG signals by K-means algorithm. The original ECG data are pre-processed at first. Then data noise reductions are done by wavelet threshold denoising method. After comparing and analyzing some common fitting functions, cubic B-spline curve is applied to extract features of ECG signal. Because the dimensions of eigenvectors are relatively high, principal component analysis is used for dimensionality reduction. At last, K-means algorithm is used to cluster the ECG signals. Results are excellent.By using VC++and MATLAB, a platform for automatic diagnosis of ECG signals has been developed. Algorithms in each step are verified for its correctness and effectiveness through this platform. The research has a realistic significance for ECG automatic diagnosis.