Efficient Extreme Learning Machine (ELM) Based Algorithm for Electrocardiogram (ECG) Heartbeat Classification

Abstract

Electrocardiogram (ECG) estimates the electric signals activity of the human heart and is extensively used for sensing heart aberrations due to ease of use and non-invasive application on human body. Human heart is a one of the vital organs of human body. In an industrial environment, heart impairments and abnormalities are attributed to the different causes including work overload, occupational and workplace stress. Cardiovascular Disease (CD) of heart refers the conditions involving different heart’s frequency deviations and are mostly ascribed to the workplace stress, fatigue and strain. Early detection of deviated heartbeats may prevent premature morbidity and unhealthy rhythms under occupational stress. The Electrocardiography (ECG) is one of the widely used diagnostic test tools that cardiologists use to diagnose heart anomalies, impairments and diseases. Various approaches have been proposed to correctly classify the ECG signals. In this study, a fast ECG classification method based on Extreme Learning Machines (ELM) algorithm is proposed to classify the frequency rhythms in heartbeat. The MIT-BIH Arrhythmia Database having recordings of 47 subjects is used in this study. Proposed ELM method is evaluated and analyzed by dividing diagnostics datasets in 60:40 train-test split ratio and findings are compared with similar studies. Results confirm the feasibility of newly proposed ELM method both in terms of classification accuracy 97.55%, speed and computational power.