| T O P I C R E V I E W |
| Sharath U |
Posted - 02/06/2012 : 10:04:07 PM
Hi,
I need to fit a curve whose data is logged at low frequency sampling rate. But i have multiple peak events in few samples, in the curve which needs to remain. Basically i just want the noise to be smoothened to show the actual peak that is recorded. I have tried the standard smoothing functions, but to no avail, as the peak will also be smoothed since the low frequency.
Pls help me out, by letting me know how to do this.... |
| 1 L A T E S T R E P L I E S (Newest First) |
| Drbobshepherd |
Posted - 02/07/2012 : 10:48:37 AM
You say you have a curve sampled at a low sampling rate and you want to find the peaks. Hopefully, the sampling rate is not so low that you failed to record the events.
If the events occur with constant periods, you might try Fourier analysis. Perform the FFT, apply a low-pass filter, then perform the inverse FFT. Before applying the final transform, try a technique called zero padding to increase the resolution of your final curve.
If your events are not periodic, try the Smoothing tool with the Savitzky-Golay filter. S-G filtering was developed for the purpose of preserving peak shapes. Experiment with various parameters to find what you think is the best fit. Tip: The larger order polynomials tend to preserve narrower features, but they will allow more hi-freq noise to pass through the filter. |
|
|
