Payam Bahman-Bijari - Control and Intelligent Processing Center of Excellence, Electrical and Computer Engineering Department, University of Tehran, Tehran, lran , School of Cognitive Sciences, Institute for Studies in Theoreticat Physics and Nlathematics (IPNI), Tehran, Iran
Alireza Akhoundi-Asl - Control and Intelligent Processing Center of Excellence, Electrical and Computer Engineering Department, University of Tehran, Tehran, lran , School of Cognitive Sciences, Institute for Studies in Theoreticat Physics and Nlathematics (IPNI), Tehran, Iran
Fariba Bahrami
Ali Jalali - Faculty of Mechanical Engineering, Khaje Nasir Toosi University of Technology.

Automatic classification of cardiac arrhythmia is a challenging area in the field of heart abnormality detection. Conventional methods used to classify arrhythmia use feature based inforntation related lo ECG signal. In this paper a novel methocl is introduced, to extract specific ntedical idormation using ECG data from leads containing this information for each arrhythmia. We have shown that using L'l in addition to VII improves the results of classification In fact, in data obtained from L'l special patterns appear which deal with Lefi Bundle Branch Block Beat (LBBB) and Right Bundle Branch Block Beat (RBBB), and this information helps medical doctors to detect arrhythmia. Adding this feature to the classification algorithm increases the accuracy while resztlting in less complex classifiers. After including the dala of the leads with accurate infonnation about each anhythmia, we reduced exlrentely the number of inputs wing a Fuzzy set-based feature extraction method. Ilavelet coefficients of the ECG signal were fed into a simple preceptron neural network consisting of one hidden layer as input Since specifc leads were used high accuracy was achieved despite the reduced number of inputs and the simplicity of the network In the present work the ECC data is taken from standard MIT-BIT Arrhythmia database

Arrhythmia, ECG, Wavelet, Neural Network