Website Based Classification of ECG Signal Using Hjorth Descriptor

Abstract— Healthiness is important to our body for living our life and work for our living. Nowadays, we encounter many health problems, but the high-rate cause of death is Heart disorder. It lives among us quietly and hard to be detected without medical diagnosing devices and experienced medical staffs. But when detectable, most cases will be in the final serious period of health problem. In this paper we address such issue and have an attempt to develop the application that can help on analysis of heart disorder. As reported in the past the effective feature that is used for screening heart disorder is Hjorth Descriptor. It represents as the measurement of ECG signal’s characteristics and used to classify three different types of ECG signals by machine learning, which are Normal Sinus Rhythm (NSR), Atrial Fibrillation (AF), and Congestive Heart Failure (CHF), respectively. As our classification results shown, the best model of classifier was found among several models in testing to be Extra Trees model with accuracy, recall, precision and F1 of 0.90, 0.90, 0.93 and 0.89, respectively. This can be concluded that our application can effectively classify the different categorized ECG samples and useful as the supplement of diagnosis of heart disorder and further development for improvement in future.

Keywords— Hjorth Descriptor, Normal Sinus Rhythm (NSR), Atrial fibrillation (AF), Congestive Heart Failure (CHF), Electrocardiography (ECG), Classification