Real-Time ECG Signal Feature Extraction and Classification using Support Vector Machine

This paper presents a reliable ECG signal analysis and classification approach using Discrete Wavelet Transform and Support Vector Machine (SVM). This methodology is made out of three phases, including ECG signal preprocessing, feature selection, and classification of ECG beats. The Discrete Wavelet Transform is utilized for signal preprocessing, de-noising, and for extracting the wavelet coefficients. These features of every ECG beat are given as inputs to the classifier. Out of the different classifiers, SVM is used to classify the input ECG beat into one of the 6 heartbeat types. Around 6400 ECG beats were chosen from the MIT-BIH arrhythmia database for this work. The best level of accuracy was obtained when level 4 approximate coefficients with Daubechies (db2) filter were used for classification. The obtained average accuracy of the SVM classifier is 98.67% using the ECG signals from the MIT-BIH database. The real-time ECG signals from different people were recorded using the HealthyPi V3 Kit, tested with the classifier model and achieved an average accuracy of 98.61%.