Signal Reconstruction from Level Crossing ADC

The processing of electrocardiogram (ECG) signals enables the diagnoses of many different heart conditions. Finding signal processing methods that reduce the computational intensity of ECG processing algorithms can facilitate faster and cheaper identification of often life threatening cardiovascular diseases. This conference paper seeks to analyse different methods of reconstructing ECG signals that have been sampled by a Level Crossing Analog to Digital Converter (LC-ADC). Traditionally, ECG signals are sampled uniformly in time. However energy savings can be achieved by using an LC-ADC at the cost of the ECG signal being non-uniformly sampled in time, making downstream signal processing more difficult. Classification of these signals often requires post-processing techniques to derive a fixed length vector of features as an input to a conventional classifier. One way to achieve this is through the use of time-domain interpolation. We will compare the results of several different interpolation algorithms with the goal of finding the most accurate method of reconstructing the signal sampled by the LC-ADC.