Help with fitting curve to data

Hey all,

I'm new to matlab, and I have to make a smoothed curve to fit the peaks of my data, similar to this picture:

http://i.imgur.com/pf4UB.png

I've been trying to use findpeaks(), like in this example: http://www.mathworks.com/help/toolbox/signal/ref/findpeaks.html

But I can't produce the curve like in the picture, any good ideas?

Thanks!

Donna

Answer 1

[bym]

Maybe one of these enveloping tools could help:

http://www.mathworks.com/matlabcentral/fileexchange/?term=envelope

however, most of the PSD envelopes I've seen are done manually, not by algorithm

Answer 2

[H. Donna Gust]

I found that I could use specplot and take the 4th power of whatever spectrum and use that to fit nicely over, example:

http://i.imgur.com/qMmXw.png

so: semilogy(x,y); hold on; specplot(spectrum(x.^4,128),Fs);

Answer 3

[H. Donna Gust]

Actually, does anyone know how I would go about collecting the actual data points of this operation? It may sound strange I guess, but I'd like to save the values of the line above the spectra data. I can't figure out how to simply get a point on that line with the corresponding frequency...any thoughts?

Donna

Answer 4

[H. Donna Gust]

Figured this out eventually...

h=figure; semilogy(x,y); hold on; specplot(spectrum(x.^4,128),Fs);

ax=findall(h,'Type','line'); xx=get(ax,'XData'); yy=get(ax,'YData'); xxx=[xx{1,1}' yy{3,1}']; % the values {1,1} and {3,1} may vary for you, i found which ones i needed by opening the fig, then -> edit -> plot browser, and selecting the lines i needed, then to properties to see the values

semilogy(xx{1,1},yy{3,1}); %just to make sure it's the same line