Assessment of resampling methodologies of electrocardiogram signals for feature extraction, statistical and neural networks applications

Developing new applications in statistics and neural networks related to ECG pattern recognition problems require large, well-tested databases. A procedure needs to be developed to normalize the sampling frequencies of ECG data from various sources to a single sampling frequency. The present work provides a methodology to compare the effectiveness of different resampling techniques. The comparison is based on the least-distortion of the original signal in time and frequency domains. Results indicate the numerical analysis methods to be superior to signal processing methods in terms of simplicity, speed and accuracy, for the cases of both high and low sampling rates. This procedure helps to speed up the data collection process for large studies.