Middle-East Journal of Scientific Research 23 (5): 896-901, 2015
ISSN 1990-9233
© IDOSI Publications, 2015
DOI: 10.5829/idosi.mejsr.2015.23.05.22226
Corresponding Author: V.R.S. Rajesh Kumar, Anna University, Chennai, Tamilnadu, India. E-mail: rajkumarresearch@gmail.com.
896
Feed Forward Neural Network Optimized Using PSO and
GSA for the Automatic Classification of Heartbeat
V.R.S. Rajesh Kumar and A. Sivanantharaja
1 2
Anna University, Chennai, Tamilnadu, India
1
Alagappa Chettiar College of Engineering and Technology, Karaikudi, Tamilnadu, India
2
Abstract: In this paper, an automatic classifier has been developed using Feed Forward Neural Network (FFNN)
to classify the ECG signals between different heartbeats. Here, the classifier is trained independently
bymorphological, heartbeat interval features and temporal features using Particle Swarm Optimization (PSO) and
Gravitational Search Algorithm (GSA). The trained classifier then classifies the beats into Normal beat (N),
Premature Ventricular Contraction (PVC), Right Bundle Branch Block Beat (R), Fusion of Paced and Normal Beat
(f), Fusion of Ventricular and Normal Beat (F) and the Atrial Premature Beat (A). The classifier performance is
validated using the benchmark database such as MIT-BIH and the performance of the classifier trained
independently using PSO and GSA is compared. It is observed that FFNN trained with GSA performs better
than the one trained with PSO.
Key words: Feed Forward Neural Network (FFNN) Particle Swarm Optimization (PSO) Gravitational
Search Algorithm (GSA) Heart beats Normal beat (N) Premature Ventricular Contraction
(PVC) Right Bundle Branch Block Beat (R) Fusion of Paced and Normal Beat (f). Fusion of
Ventricular and Normal Beat (F) and Atrial Premature Beat (A)
INTRODUCTION Markov models, linear regression, support vector machine
Heartbeat classification is a significant step in the the classification of heart beats. Among these techniques
diagnosis of arrhythmia [1, 2]. An arrhythmia is the sign classifications using neural network provides promising
of irregular heartbeat, which may be too fast or too slow results.
or inconsistent. A normal person’s heart rate lies in the Thus, in this paper a Feed Forward Neural Network
range of 60-100 beats per minute (BPM). If the rate is less (FFNN) has been employed to classify the ECG signals
than the normal rate then it is the indication of into six beats: Normal beat (N), Premature Ventricular
bradycardia and tachycardia if the rate is larger than the Contraction (PVC), Right Bundle Branch Block Beat (R),
normal rate [3]. Various techniques such as Fusion of Paced and Normal Beat (f). Fusion of Ventricular
electrocardiogram (ECG), Holter Monitoring, Event and Normal Beat (F) and Atrial Premature Beat (A). Here
Recorder, Chest X-ray and so on are utilized to analyze the FFNN is trained independently using PSO and GSA
arrhythmias, among theseutilizing ECG is the general techniques with temporal, heartbeat intervals and
approach.Traditionally, physicians analyze the cardiac morphological features of the heart beat in the signal. The
activity via ECG signal manually and determine the extracted features are fed into FNN to classify the
presence of arrhythmia, which is a time consuming different types of beats.
process. The rest of the paper is organized as follows: section
In recent years, computerized technique has been II presents the research work that has been carried out in
used for the automatic classification of ECG signals. order to classify the heartbeats, section III describes the
Several machine learning techniques such as expert proposed approach, results are discussed in section IV
systems, heuristic approaches, self-organizing map, and finally section V concludes the paper.
and neural networks and so on have been employed for