ECG Arrhythmia Classification Using Evolved Multilayer Perceptron Neural Network

This paper presents evolvable multilayer perceptron neural network (MLPNN) for electrocardiogram heartbeat classification based on a combination of morphological and temporal features. Data has been obtained from the MIT-BIH database to classify heartbeats to one of the five beat classes recommended by AAMI standard. For classification of the ECG signals, a hybrid training algorithm has been used and MLPNN weights have been optimized using genetic algorithm. Then back-propagation algorithm has been used as a local optimization operator. The main advantage of weight evolution by genetic algorithm is to simulate the learning process of a neural network, avoiding the drawbacks of the traditional gradient descent, such as back-propagation. Simulation results demonstrate high average detection accuracy of ECG signal patterns.

Keywords:

propagation , Electrocardiogram (ECG) , genetic algorithm , QRS complex , multilayer perceptron