A QRS Complex Detection in Multi-lead ECG signals Based on Filtering and Standard Deviation Shift

One of the most vital parts of the electrocardiogram (ECG) is the QRS complex. It is considered as an essential basis for determination of heart condition. Since the nature of an ECG signal is non stationary, the accuracy of detecting R -peaks is decreased which may result in substantial increase in false negatives or positives. In this paper, we propose a multi-lead QRS detection algorithm. A certain random number of leads are input to the algorithm which is based on filtering. However, each lead is applied to a stationary wavelet transform and two median filters to reduce the effect of other parts in the ECG signal. A smoothing filter is applied reducing the amplitudes of unwanted high frequency content. In order to further reduce the amplitudes of the P and T waves, the ECG signal is shifted towards the negative part of the Y-axis proportional to its average standard deviation. The QRS candidates of each lead are arranged in vectors and input to a combining algorithm which is based on voting. The algorithm proposed in this paper assures acceptable detection rate and serves as an active basis for various ECG processing. The performance of the proposed algorithm is evaluated on the St. Petersburg INCART database which is a 12-lead database. Experimental results show that a sensitivity of 99.828% and detection error rate of 0.636% is achieved.