Fast R-Peak Detection from Compressed ECG Sensing Measurements Without Reconstruction for Energy-Constrained Cardiac Health Monitoring

For resource-constrained health monitoring devices, we propose a fast R-peak detection from compressed electrocardiogram (ECG) sensing measurements without performing ECG reconstruction. A suitable sensing matrix is presented to generate CS measurements, which can be directly used for detecting R-peaks. The proposed method includes measurement generation, Gaussian derivative filter, fixed amplitude thresholding, Shannon energy based peak amplification, envelope differencing and negative zerocrossing determination for locating R peaks of candidate smoothed QRS complex and R peak correction. The proposed method is evaluated using the standard MIT-BIH database and had an overall accuracy of 99.47%, sensitivity of 99.65%, and a positive predictivity of 99.82%. The proposed method outperforms other existing QRS detection methods in terms of processing time (computational load reduction by 75%) and comparable accuracy without using sets of thresholds, which are conflict in practice for the ECG signals having time-varying heart rates, low- and wide-QRS complexes.