Publication

Abstract : Cardiac disease detection is a tedious process. Classification of electrocardiogram (ECG) signals plays an important role in the diagnosis of heart diseases. The most important factor that limits the detection of cardiac disease is the rare availability of instances of the abnormal condition collected using ECG sensors. And if the signals contain noise, then the classification might become a challenging task. In this work, we address the problem of cardiac disease detection when the dataset has less number of noisy ECG sensor signals. Here, Chebyshev Type II filter and Chebyshev function, which is termed as Chebfun, are used. The Chebyshev filter is used for high-frequency noise removal and Chebfun is used to approximate the signal with its coefficients. These coefficients known as Chebfun coefficients are used as the features. These features are used for classification. In the proposed work, machine learning algorithms, like SVM, logistic regression, decision tree, and AdaBoost, are used for classifying the features extracted from Chebfun.