# Determine 3 most dominant frequencies in a signal, PSD or just the absolute value of a Fourier transform?

I have a noisy ECG signal recorded for 5 minutes. My goal is to determine heart rate every 2 seconds. To find out 3 most dominant heart rates (Beats per minute) in a 2s-signal should I calculate its PSD or just the absolute value of a Fourier transform?

• Why not just find the times between the R peaks of the QRS complexes and compute the frequency using those time interval measurements? That's going to be more precise than FFT bins which have a finite width, and hence ambiguity. – Andy Walls Nov 29 '18 at 18:13
• Because the signal is too distorted caused by motion artefact. – Ηλεκτρολόγος Μηχανικός Nov 29 '18 at 20:01
• The ECG is chest strap type ECG. – Ηλεκτρολόγος Μηχανικός Nov 29 '18 at 20:03
• 60 bpm is 1 Hz, 120 bpm is 2 Hz. Given your sample rate and planned number of FFT points, how many Hz per bin of frequency resolution are you going to have? At 60 bpm, 2 seconds of data is only 2, maybe 3, heartbeats. What window are you going to use on the time domain data before the FFT, and will it destroy one of the beats in the measurement interval? – Andy Walls Nov 29 '18 at 21:45