Hints to smoothe noisy data

Hi can someone of you give me a tip how i could best smooth following data to eliminate noise and to emphasize the peaks. I attached the plot of the raw measurement beow.

[Adam Danz]

Here's the figure so people don't have to download it:

[Adam Danz]

What do you consider a peak? What would your ideal plot look like after smoothing?

Answer 1

[Cris LaPierre]

"Best" way depends on many things. Classic ways to remove noise are signal averaging if your signal is repeatable, or applying a moving average. If you know something about the frequency of your desired signal, you could apply a filter.

Take a look at this example page.

Answer 2

[Bjorn Gustavsson]

It sounds like you want a filter that is similar to what you get with the wiener2 function - from the top of my head I dont remember if it works just as great on 1-D signals, but you can look at its source-code. This filter (Lee's sigma-filter) filters variations that are small compared to the local variance by keeps spikes.

HTH