Impedance Cardiography heartbeat classification using LP, DWT, KNN and SVM

In this work, a computer aided diagnosis system is proposed to classify the Impedance Cardiography signals ‘ICG’ into two groups which are normal and abnormal. The ICG signals are denoised by using the discrete wavelet transform DWT ‘db8’ in order to eliminate different kinds of artifacts. Then, each ICG signal is decomposed into several heartbeat segments by using the location of C peaks. Furthermore, the Linear Prediction model ‘LP’ and the Discrete Wavelet Transform ‘DWT’ are used to extract temporal and time-frequency features, respectively. Total of 21 extracted features are selected and used to classify the ICG heartbeat segments. Besides, the K-Nearest Neighbor ‘KNN’ and the Support Vector Machine ‘SVM’ classifiers are evaluated and their performances are compared to find the approach that gives the best classification results. The proposed method achieves high accuracy of 100% when using the db8 wavelet to extract features and the SVM as a classifier.