The Research Of J Wave Detection Technique Based On Multi-feature Recognition

J wave is the hump-shaped,slurring or spike-shaped waveform at the end of QRS complex on the electrocardiogram(ECG).J wave can indicate the appearance of malignant arrhythmia,sudden cardiac death and other cardiovascular diseases and be a warning index for the detection of certain cardiac disease.It is necessary to find an efficient and accurate J wave detection technique to provide important basis for the clinical diagnosis of J wave related diseases.This thesis presents two J wave detection methods,the specific is as follows:The first J wave detection method is based on time-frequency domain features and adaptive feature selection.Three sets of features are extracted from normal ECG data and ECG data with J wave in the time-frequency domain after preprocessing,which include morphological features,the statistical features based on Intrinsic Mode Function and Hilbert-Huang Transform.A feature selection method based on distinction degree is proposed to reduce the dimension of the extracted feature set.Support Vector Machine is trained by the selected feature set and its parameters are optimized by Particle Swarm Optimization algorithm.Finally,the classification model is applied to predict and identify the test data.The second J wave detection method is based on the blended feature extraction.The R+75 and its statistical features in the time domain,Discrete Cosine Transform features in the frequency domain,and two nonlinear high order statistical features are extracted from the two types of ECG data.Two kinds of nonlinear features are reduced dimensionally by Linear Discriminant Analysis.Feature fusion is performed in features extracted from time domain and frequency domain and nonlinear features after dimension reduction.Probabilistic Neural Network is used to complete classification and identification and the optimal value of influencing factor is determined by data analysis.The experimental results show that the two methods proposed in this thesis both have high recognition rate of J wave and the average accuracy are 94.5% and 95.0% respectively,which means J wave can be detected accurately from ECG data.