Abstract— Among all ECG components, QRS complex is the most significant feature. Entropy based method for the detection of QRS complexes (cardiac beat) in the single lead Electrocardiogram (ECG) is proposed in this paper. Digital filtering techniques are used to remove noise and base line wander in the ECG signal. Entropy criterion is used to enhance the QRS complexes. Support Vector Machine (SVM) is used as a classifier to delineate QRS and nonQRS regions. The performance of the algorithm is evaluated against the standard CSE ECG database. The numerical results indicated that the algorithm achieved 99.68% of the detection rate. The percentage of false positive and false negative is 2.28 and 0.32 respectively. The detection rate depends strongly on the quality of training, data representation and the mathematical basis of the classifier. Index Terms— ECG, Entropy, QRS complex, SVM. I. INTRODUCTION Electrocardiogram (ECG) provides useful information about functional status of the heart. Analysis of ECG is of great importance in the detection of cardiac anomalies. In a clinical setting, such as intensive care units, it is essential for automated systems to accurately detect and classify electrocardiographic signals. As displayed in Fig. 1, ECG is characterized by a recurrent wave sequence P, QRS and T associated with each beat. The QRS complex is the most striking waveform, caused by ventricular depolarization of the human heart. Once the positions of the QRS complexes are found, the detection of other components of ECG like P, T waves and ST segment etc. are found relative to the position of QRS, in order to analyze the complete cardiac period. In this sense, QRS detection provides the fundamental for almost all automated ECG analysis algorithms. Manuscript received May 3, 2007. This work was supported by All India Council for Technical Education (AICTE), N. Delhi, India. S. S. Mehta is with the Department of Electrical Engineering, MBM Engineering College, J. N. V. University. Jodhpur- 342 001, Rajasthan (India). (e-mail: ssmehta_58@rediffmail.com) N. S. Lingayat is with Department of Electrical Engineering, Institute of Petrochemical Engineering, Dr. B. A. Technological University, Lonere-402 103, Maharashtra (India). Presently he is on deputation as a Research Scholar at Department of Electrical Engineering, MBM Engineering College, J. N. V. University. Jodhpur- 342 001, Rajasthan (India). (phone: +91 291 2515488, fax: +91 291 2513348, e-mail: nslingayat@yahoo.com). Fig.1 ECG Signal Numerous QRS detection algorithms such as derivative based algorithms, algorithms based on digital filters, wavelet transform, length and energy transform, artificial neural networks, genetic algorithm, syntactic methods, Hilbert transform etc. are reported in literature. Kohler et al [1] described and compared the performance of all these QRS detectors. Recently, few other methods based on pattern recognition [2], Hilbert transform [3], wavelet transform [4], neuro-fuzzy approach [5], filtering technique [6], first derivative [7], curve length concept [8], moving-averaging incorporating with wavelet denoising [9] etc. are proposed for the detection of QRS complexes. Christov et al [10] gave a comparative study of morphological and time-frequency ECG descriptors for heartbeat classification. Most of these QRS detectors are one channel detectors. A common technique utilized in the QRS detector algorithm is to employ a scheme that consists of a preprocessor and a decision rule [11]. The purpose of the preprocessor is to enhance the QRS, while suppressing the other complexes as well as the noise and the artifacts. The preprocessor consists of a linear filter and a transformation. The purpose of the decision rule is to determine whether or not QRS complex is present at a given instant in the signal. SVMs based classification methods represents a major development in pattern recognition research. Two innovations of SVMs are responsible for the success of these methods, namely, the ability to find a hyperplane that divides samples in to two classes with the widest margin between them, and the extension of this concept to a higher dimensional setting using kernel function to represent a similarity measure on that setting. Both innovations can be formulated in a quadratic programming framework whose optimum solution is obtained in a computation Support Vector Machine for Cardiac Beat Detection in Single Lead Electrocardiogram S. S. Mehta, and N. S. Lingayat, Member, IAENG IAENG International Journal of Applied Mathematics, 36:2, IJAM_36_2_4 ______________________________________________________________________________________ (Advance online publication: 24 May 2007)