ECG Feature Extraction by Multi Resolution Wavelet Analysis based Selective Coefficient Method

One of the major problems in feature extraction methodologies from ECG is to identify the wave and complex boundaries. Here we present a multiresolution wavelet based approach to identify the boundaries. The wavelet reconstruction coefficients are assessed in terms of shape and size to eliminate the interfering components for better detection of wave boundaries. The relevant coefficients are retained that resembles with the original structure of the wave under investigation. The algorithm suggests different set of reconstruction coefficients for different part of the ECG wave which eliminates the scope of probable interaction between adjacent regions and thus correct identification of wave boundaries are ensured. The QRS complex and T wave duration are measured with the algorithm and validated against some arbitrarily chosen ECG records for lead 1 from Physionet PTB diagnostic database. The measured values are compared with the manually determined values and the accuracy for each evaluation is calculated. The test result shows over 95% accuracy for QRS complex and over 92% accuracy for QT interval and T duration.