I S SIVA RAO et al.: QRS DETECTION OF ECG - A STATISTICAL ANALYSIS 1080 QRS DETECTION OF ECG - A STATISTICAL ANALYSIS I.S. Siva Rao 1 , T. Srinivasa Rao 2 and P.H.S. Tejo Murthy 3 1 Department of Information Technology, Raghu Engineering College, India E-mail: isro75@gmail.com 2 Department of Computer Science and Engineering, GITAM Institute of Technology, GITAM University, India E-mail: tsr.etl@gmail.com 3 Department of Electrical and Instrumentation Engineering, GITAM Institute of Technology, GITAM University, India E-mail: teja_usha@rediffmail.com Abstract Electrocardiogram (ECG) is a graphical representation generated by heart muscle. ECG plays an important role in diagnosis and monitoring of heart’s condition. The real time analyzer based on filtering, beat recognition, clustering, classification of signal with maximum few seconds delay can be done to recognize the life threatening arrhythmia. ECG signal examines and study of anatomic and physiologic facets of the entire cardiac muscle. The inceptive task for proficient scrutiny is the expulsion of noise. It is attained by the use of wavelet transform analysis. Wavelets yield temporal and spectral information concurrently and offer stretchability with a possibility of wavelet functions of different properties. This paper is concerned with the extraction of QRS complexes of ECG signals using Discrete Wavelet Transform based algorithms aided with MATLAB. By removing the inconsistent wavelet transform coefficient, denoising is done in ECG signal. In continuation, QRS complexes are identified and in which each peak can be utilized to discover the peak of separate waves like P and T with their derivatives. Here we put forth a new combinatory algorithm builded on using Pan-Tompkins' method and multi-wavelet transform. Keywords: Electrocardiogram (ECG), QRS Detection, Wavelet Transform, Denoising, Pan-Tompkins’ 1. INTRODUCTION The most critical part of the electrocardiogram (ECG) is the QRS complex. Its shape and occurrence time accord much statistics and details about the heart function. Since it has a distinguished shape, QRS detection is the principal part of almost all automated ECG analysis algorithms such as heart rate variability and cardiac cycle classification. A typical ECG waveform is shown in Fig.1. A prototypical QRS detection algorithm generally consists of two phases: preprocessing and decision [1].Comprehensively, the former includes some sort of filtering [2], while the latter attempts to specify the location of QRS complexes in the ECG signal. Till date, an extensive variety of techniques such as linear filtering, neural networks [5], mathematical morphology and wavelet transforms have been recommended by the researchers of QRS detection. For instance, a substantial number of QRS detection algorithms principally use linear filtering to remove objectionable parts of the ECG signal. Then, QRS complexes are determined by applying a suitable threshold to the resultant signal. Pan-Tompkins' method [2], [3], is a familiar algorithm of this classification. Offlate, a number of QRS detection algorithms were suggested based on the wavelet transform. They examine the ECG signal by using the multi-scale sub-band data and statistics provided by the wavelet transform. Fig.1. Normal ECG Waveform In addition, a number of combinatory algorithms were earlier proposed, to be benefitted of the linear filtering and wavelet transform. Predominantly, QRS detection is a not a minor task. It may possibly be malformed by noise caused by the electrode artifact, baseline drift, and power line interference [9]. Almost always, the ECG signal may deform owing to the pathological or obsessive variations, e.g. signals with small QRS complexes or suddenly variable levels. Hence, to provide a dependable and a well-grounded QRS detection algorithm is still an unsolvable open problem. In this paper, a novel hybrid algorithm is proposed which integrates various wavelet coefficients and Pan-Tompkins' method for extracting features [10] and also improved the computational cost by using different parameters like uniformity, entropy, etc., to analyze ECG signal statistically. 2. HEART DISEASES The term “Heart Disease” refers to conditions that involve narrowed or blocked blood vessels that can lead to a heart attack, chest pain (angina) or stroke. There are different types of heart diseases, the most common type that affects the electrical system is known as arrhythmias. They can cause the heart to beat very fast (Tachycardia) or very slow (Bradycardia), or unexpectedly (Atrial fibrillation).