Sounds like a good project.
Steve is right. It can be done more easily and with good results in the time domain, not frequency domain. You'll want to apply a bandpass filter, f_low=1 Hz, f_high=40 Hz, to remove drift and high frequency noise. I recommend Zero Phase Filter.vi due to its nice phase properties. If you use it, you must also use the Butterworth Coefficients.vi followed by Cascade to Direct.vi, which prvides the filters coefficients in the correct (forward and reverse) format that Zero Phase Filter.vi needs. Or you can just use Butterworth.vi, instead of Zero Phase plus Butterworth Coefficients plus Cascade to Direct. Whether you use Butterworth,vi or Butterworth Coefficients.vi, you want to specify 2nd order bandpass with 1 and 40 Hz cutoffs. Now take the filtered ECG and peak-detect the R waves, as Steve said. Oeak Detector.vi is good. It puts out the number of peaks found. Divide by the signal duration and you have events per unit time. Convert time from events/second to events/minute since you want beats per minute.
I have tried differentiating ECGs, on the theory that the sharp rise of the R wave would be even easier to detect than the peak of the R-wave. In practice, I found this didn't work as well as just using the un-differentiated ECG. Plot the bandpass-filtered ECG. Add a horizontal cursor to the graph, which will serve as the inidcator of the detection threshold. Add a property node to get the current cursor height. Use that height as the threshold for Peak Detector.vi (under the Signal Processing palette). The user sets the detection threshold by dragging the cursor up or down. Put the graph and cursor-reading and peak-detecting and HR-calculating inside a while loop so the user can experiment with different cursor settings and observe the effects on computed heart rate. Keep the filtering outside the while loop since it only needs to be done once.
