Multiple Channel Electrocardiogram QRS Detection by Temporal Pattern Search

A highly constrained temporal pattern search ("TEPS") based multiple channel heartbeat detector is described. TEPS generates sequences of peaks and statistically scores them according to: 1) peak time coherence across channels; 2) peak prominence; 3) temporal regularity; and 4) number of skipped beats. TEPS was tested on 31 records of three channel capacitive electrode data from the UnoViS automobile database. TEPS showed a sensitivity (SE) of 91.3% and a false discovery rate (FDR) of 3.0% compared to an SE and FDR of 75.3% and 65.0% respectively for a conventional single channel detector (OSEA) applied separately to the three channels. The peak matching window was 30ms. The percentage of 5 second segments with average heart rates within 5 beats/minute of reference was also measured. In 6 of the 31 records, TEPS percentage was at least 30% greater than OSEAs. TEPS was also applied to synthetic data comprising a known signal corrupted with calibrated amounts of noise. At a fixed SE of 85%, increasing the number of channels from one to two resulted in an improvement of approximately 5dB in noise resistance, while increasing the number of channels from two to four resulted in an improvement of approximately 3dB in noise resistance. The quantification of noise resistance as a function of the number of channels could help guide the development of wearable electrocardiogram monitors.