Support Vector Machine-Based Expert System for Reliable Heartbeat Recognition

Abstract

This paper presents a new solution to the expert system for reliable heartbeat recognition. The recognition system uses the support vector machine (SVM) working in the classification mode. Two different preprocessing methods for generation of features are applied. One method involves the higher order statistics (HOS) while the second the Hermite characterization of QRS complex of the registered electrocardiogram (ECG) waveform. Combining the SVM network with these preprocessing methods yields two neural classifiers, which have been combined into one final expert system. The combination of classifiers utilizes the least mean square method to optimize the weights of the weighted voting integrating scheme. The results of the performed numerical experiments for the recognition of 13 heart rhythm types on the basis of ECG waveforms confirmed the reliability and advantage of the proposed approach.

Multiple neural network integration using a binary decision tree to improve the ECG signal recognition accuracy

“…The structure of the MLP was selected by the trial-and-test method, in which we started with a network with a small number of hidden neurons and we increased the hidden neurons until a good testing error was achieved. The SVM learning parameters were selected as in the work of Osowski et al (2004). With 7 classes and the one-against-one method to find the winner class in the SVM, a total number of 21 SVMs were trained.…”

“…In our approach to the problem, we applied the QRS complex decomposition into Hermite basis functions and used the decomposition coefficients as the features of the ECG signals. These coefficients, together with two classical time-based features: the instantaneous R-R interval of the beat (the time span between two consecutive R peaks) and the average R-R interval of 10 preceding beats, form the feature vector x applied to the input of the classifier (Osowski and Linh, 2003;Osowski et al, 2004;2006). In the Hermite basis function expansion method, we have Hermite polynomials defined by a recurrent formula:…”

“…The most distinctive fact about the SVM is that the learning task is simplified to quadratic programming by introducing the Lagrange multipliers α i . All operations in learning and testing modes are done in the SVM using kernel functions K(x, x i ), satisfying the Mercer conditions (Osowski and Linh, 2003;Osowski et al, 2004). The best known kernels are Gaussian, polynomial, linear or spline functions.…”

Detecting premature ventricular contractions in ECG signals with Gaussian processes

“…For such purpose, simple as well as sophisticated algorithms exploiting different classification strategies with different features representations of the ECG signals have been proposed [1]- [3].…”

“…In addition, we report an experimental comparison of different kinds of ECG signal representations, which are the standard temporal signal morphology, the discrete wavelet transform domain [1], the S-transform [7], which is an extension to the ideas of wavelet transform, and the highorder statistics [3]. The main idea of GPs is to assume that the probability of belonging to a class label for an input beat is monotonically related to the value of some latent function at that beat.…”