How can I find the peaks of a cfit object?

Hi, I am trying to find the peaks of a fitted curve that I have got using the curve fitting toolbox. I have a fitting function that is doing this:

 ft = fittype( 'smoothingspline' );
opts = fitoptions( 'Method', 'SmoothingSpline' );
opts.SmoothingParam = 1.5029271581647606E-4;
fitresult, gof] = fit( xData, yData, ft, opts );

And a main function that I'm trying to get working using something like this

[fit,gof] = findfit(Z2);
test = coeffvalues(fit);
peaks = findpeaks(test.coefs);

with an error: "Expected X to be a vector"

I think my question boils down to, how do I use something like findpeaks() on a cfit object that I get returned from my graphing function.

Thanks for all your help.

Edit: It might be worthwhile to know Z2 3x295(3 principle components per column), test.coefs 884 x 4. The purpose of using the smoothed curve is that prior to this the curve was too 'jagged' and so I would not be able to build something that easily finds the peaks & troughs of a signal. Apologies if this is simple, I'm a beginner.

Answer 1

[Matt Tearle]

test.coefs is giving you the coefficients of the cubic splines that make up the fit. If I interpret your intention correctly, you're wanting the actual fitted curve. To do that, use feval:

yFitted = feval(fit,xData);

Now you can use findpeaks on yFitted (assuming that yData was a vector originally, so this is a 1-D fit).

xData = sort(rand(295,1));
yData = sin(6*pi*xData) + 0.1*randn(295,1);

ft = fittype( 'smoothingspline' );
opts = fitoptions( 'Method', 'SmoothingSpline' );
opts.SmoothingParam = 0.9999;
[fitresult, gof] = fit( xData, yData, ft, opts );

yfitted = feval(fitresult,xData);

plot(xData,yData,xData,yfitted)

[ypk,idx] = findpeaks(yfitted);

xpk = xData(idx);

hold on
plot(xpk,ypk,'o')

[Chenhan]

This is really helpful and saves me loads of time.

Thank you very much!

Answer 2

[Montgomery]

Hi Matt, that looks great but I'm still a little confused. Currently in main I have

[fit,gof] = findfit(Z2);

and findfit() is this:

        function [fitresult, gof] = find(Z2)
        [xData, yData] = prepareCurveData( [], Z2 );
        ft = fittype( 'smoothingspline' );
        opts = fitoptions( 'Method', 'SmoothingSpline' );
        opts.Smoothing
        [fitresult, gof] = fit( xData, yData, ft, opts );

Are you suggesting that I replace my current fitfunction with the one you gave above?

Also what does this do?

yData = sin(6*pi*xData) + 0.1*randn(295,1);

Sorry If I'm being slow, its 10.20PM in my time, Thanks for taking the time to respond

[Matt Tearle]

My code was just as an example to show how feval would work. The lines xData = ... and yData = ... are just making some example data (since I don't have access to yours).

Your function is mostly fine, except that you define xData in there as a local variable, and you need that for feval. So you may want to modify findfit to something like this:

function [fitresult, gof,yFitted] = find(Z2)
[xData, yData] = prepareCurveData( [], Z2 );
ft = fittype( 'smoothingspline' );
opts = fitoptions( 'Method', 'SmoothingSpline' );
opts.SmoothingParam = 1.5029271581647606E-4;
[fitresult, gof] = fit( xData, yData, ft, opts );
yFitted = feval(fitresult,xData);

Then in your calling code,

[fit,gof,yFit] = findfit(Z2);
peaks = findpeaks(yFit);

Answer 3

[Montgomery]

Got it working in the end, thanks