I suspect the general approach would be similar to other biomedical signal processing. Calculate the Fourier transform of your data to see if the noise is broadband or band-limited. If band-limited, use a frequency-selective filter. If broadband, the options are limited. Use wavelet denoising or the Savtitzky-Golay filter to deal with it.
I have a difficult time understanding how heart rate data could be noisy, since ideally it should be derived from EKG R-R intervals. The only valid R-deflections to usee for this are those preceded by a normal P-deflection, ‘normal’ being defined in terms of timing and morphology.
