Abstract—There is an increase in cardio logical patients all over the world due to change in modern life style. It forces the medical researchers to search for smart devices that can diag- nosis and predict the onset of cardiac problem before it is too late. This motivates the authors to predict Arrhythmia that can help both the patients and the medical practitioners for better healthcare services. The proposed method uses the frequency domain information which can represent the ECG signals of Arrhythmia patients better. Features representing the MIT- BIH Arrhythmia are extracted using the efficient Short Time Fourier Transform and the Wavelet transform. A comparison of these features is made with that of normal human being us- ing Neural Network based classifier. Wavelet based features has shown an improvement of Accuracy over that of STFT fea- tures in classifying Arrhythmia as our results reveal. A Mean Square Error (MSE) of with wavelet transform has validated our results. Keyword—Feature Extraction; Short Time Fourier Trans- form; Wavelet Transform; Classification; Neural Network I. INTRODUCTION HE WORLD is fast expanding day by day that leads to rapid change in human lifestyle. It affects the human being both physically and psychologically. People are de- manding more logical, compact, cost effective and accessi- ble devices that can take care and guide their health regu- larly at ease. This has made the healthcare domain both bur- dening and competitive. Among many fatal diseases, the heart care is one of the mostly sought biomedical fields of research today. The WHO (World Health Organization) has reported around 17 million deaths only because of heart at- tack [1]. Most of the cardiovascular problems occur due to age, angina, high cholesterol levels, diabetes, diet, genetics, hypertension, smoking, HIV, work stress etc. Due to many reasons, it is very difficult to care and maintain a healthy heart always that makes the research challenging. Search for an efficient diagnosis and monitoring system that can predict accurately heart ailment has been ever increasing. It desires the automation of medical detection system that benefits the society struggling with heart related issues [2]. An effective diagnosis machine requires a signal that represents the heart disease accurately. The Electrocardiograms (ECG) has been a vital non-intrusive apparatus that provides the desired and dependable signal to the cardiologists as well as medical re- searchers. It has been a tool for analysis of different cardio- vascular arrhythmias appearing briefly during routine check of a cardiac patient [3]. T For the diagnosis and monitoring to be effective, the sys- tem desired equally discriminative features that can repre- sent the ECG signal adequately. There have been many reli- able features explored earlier by biomedical researchers in diagnosis of Arrhythmia patients using ECG. Most of these features are: temporal intervals [4], morphological features [5], statistical features [6], wavelet transform [7] etc. The ECG signal is generally described by five prominent peaks P, Q, R, S and T as given in Fig.1. Fig.1. Five Peaks of a normal ECG Signal The QRS envelope plays a major role in detection and analysis of ECG signal automatically. Wavelet transform (WT) provides the signal compression similar to STFT ex- clusively and can be utilized efficiently for normal or abnor- mal heart rhythm observation [8-9]. However, the features must be capable to simulate the pattern recognition tool for effective classification of the disease. Neural Networks (NN) have been serving as the powerful tools for data mod- eling in the past in the field of speech, image and bio-medi- cal engineering. These information processing technique can capture complex input and output relation better because of their resemblance to human brain. These networks can an- swer those problems that have no algorithmic solutions. Among a host of NNs used in pattern identification the Mul- tilayer Perceptron (MLP) is most popular and simple. It uses the hyperplanes for data space division which is the most natural approach based on the fundamental simplicity of lines and is intuitively appealing in our case. This has been used in this work for classification of the patients suffering from Arrhythmia with the chosen feature sets of ECG. Detection of Arrhythmia using Neural Network Saumendra Kumar Mohapatra, Hemanta Kumar Palo, Mihir Narayan Mohanty Electronics & Communication Engineering, ITER, Siksha 'O' Anusandhan Deemed to be University, Bhubaneswar, India {saumendramohapatra, hematapalo, mihirmohanty}@soauniversity.ac.in Proceedings of the First International Conference on Information Technology and Knowledge Management pp. 97–100 DOI: 10.15439/2018KM42 ISSN 2300-5963 ACSIS, Vol. 14 c  PTI, 2018 97