Adaptive ECG Sampling – a Minimum-Error Approach

Adaptive sampling, also known as arbitrary or compressed sensing is an interesting form of concise representation of the process in a series of samples nonuniformly distributed in time.When fed with some a priori knowledge of the signal (e.g.physiological limitation) it may provide efficient usage of the data carrier, improve precise representation of selected phenomena and increased immunity to noise.In this paper we propose a smart iterative downsampling procedure to optimize sampling timepoints for best signal fidelity at given sampling rate.The method starts with identifying and sorting local extrema and then iterates the interpolation process taking alternately one half of samples as reference and minimizing the reconstruction error by adjusting timepoints' positions of the other half.The algorithm was tested with 41 records from MIT-BIH Arrhythmia Database.We found that even with 20 sps the R-peak time precision is sufficient for accurate heart rate measurement and with only 80 sps the signal reconstruction error drops below 0.1% of the full measurement scale.