Abstract: Advanced electrocardiogram analytic applications, such as heart rate variability analysis, generally require accurate detection of R peaks and QRS complexes, simultaneously, but the existing methods are not accurate enough under interference. To address the problem, we propose a method to detect the accurate positions of QRS complexes and R peaks. The proposed algorithm begins with a non-linear transformation to enhance QRS complexes in electrocardiogram signal, and dynamic thresholds are calculated based on the detected QRS complexes to detect the new QRS complexes, and finally, an accurate R-peak identifier function is used to locate R peaks. Compared with the existing algorithms, the proposed algorithm achieves a better performance on several open-source electrocardiogram databases. It achieves 99.98% sensitivity, 99.94% positive predictivity and 0.08% detection error rate on QT Database.