Rexy J1,Velmani P 2,Rajakumar T.C3
1Manonmanium Sundaranar University,India,2The M.D.T Hindu College,India,3St. Xavier’s College,India
Heart disease is the major cause of death ratio increase in this decade. Nowadays various people of different age sector undergo the high risk of heart problems and miss their precious life all of a sudden. Early detection of heart disease will save many people’s life well in advance. Heart Diseases are predictable and they can be identified in earlier stage. First basic method to identify heart disease is ElectroCardioGram (ECG) which is the basic recording method of electrical activities of a functioning heart. ECG is the cheapest and painless method to detect the basic heart problems. This paper is an attempt to detect and classify heart beat signals which will serve as the basic step to predict basic and serious issues which may affect the functioning of the heart. The raw ECG signals are extracted and preprocessed to remove unwanted noises which will produce effective results. The preprocessed ECG signals are then are utilized to identify the heart beats which comprise of signals such as P,Q,R,S,T and U. After detecting the heart beats, they are segmented to extract the ECG Features. The temporal and spectral features are extracted from the segmented ECG signals for classification purpose. The extracted feature vectors are utilized to classify the signals. Radial Basis function and Random Forest method are commonly used classification methodologies; hence these two methodologies are applied to classify the ECG Signals into five basic classes. Massachusetts Institute of Technology-Beth Israel Hospital (MIT-BIH) Arrhythmia database and Noise Stress database are used for this implementation and the classes are identified based on the given dataset parameters. Performance metrics such as accuracy, specificity and sensitivity are computed to find out the best classification methodology among the applied two methodologies. This performance analysis provides a clear comparative view of both the existing methodologies and specifies that Radial Basis function well suits for the given segmented ECG signals and the extracted features. Hence this performance evaluation paves way for best classification algorithm selection or extension of the best methodology and it can be further optimized for better classification result. The implementation process has been carried out using Matlab software environment.
ECG, ECG Features, radial basis function, random forest method.
