Sine curve for a given period value.

I want to fit a sine curce to my data for a given period value. How should I do it?

Answer 1

[Star Strider]

See for example: Curve fitting to a sinusoidal function (link), and How to filter noise from time-frequency data and find natural frequency of a cantilever? (link). You can easily adapt the code in those to your data if you need to.

[Giru Mishra]

Hey Star Strider, I am new to MATLAB. COuld you please help me.

I have x and y data and y error..and I know its period.. how to fit a sine train..please help

[Giru Mishra]

I tried to understand through the following link:

 https://in.mathworks.com/matlabcentral/answers/121579-curve-fitting-to-a-sinusoidal-function

but could not understand because running it is giving errors. X axis of my data is phases while y axis is the data..

[Star Strider]

‘x and y data and y error.’

My code calculates the period, as well as the phase and y-offset. You simply need to give it the x vector and y vector.

I have no idea what ‘y error’ is, however it may not be possible to use it with my code.

‘... it is giving errors’

Since I do not know what the errors are, I cannot suggest a solution.

[Giru Mishra]

OKay ignore "error".

I already know period, my data's x axis is phases from 0 to 2.. and I have some y data.. now I want to fit the sine curve.

Using your code I am not able to fit.. what changes should I make according to my data?

[Star Strider]

I have no idea.

I do not have your data, and I do not know what problems you are having.

[Giru Mishra]

This is the sample data file. Please ignore 3rd column

[Star Strider]

Naming your matrix ‘data’, this works:

x = data(:,1);
y = data(:,2);

yu = max(y);
yl = min(y);
yr = (yu-yl);                                                   % Range of ‘y’
yz = y-yu+(yr/2);
zx = x(yz(:) .* circshift(yz(:),[1 0]) <= 0);                   % Find zero-crossings
per = 2*mean(diff(zx));                                         % Estimate period
ym = mean(y);                                                   % Estimate offset

fit = @(b,x)  b(1).*(sin(2*pi*x./b(2) + 2*pi/b(3))) + b(4);     % Function to fit
fcn = @(b) sum((fit(b,x) - y).^2);                              % Least-Squares cost function
s = fminsearch(fcn, [yr;  per;  -1;  ym])                       % Minimise Least-Squares

xp = linspace(min(x),max(x));

figure(1)
plot(x,y,'pb',  xp,fit(s,xp), 'r')
grid

fprintf(1, 'Amplitude = %7.3f\nFrequency = %7.3f\nPhase     = %7.3f\nOffset    = %7.3f\n\n', s(1), 1/s(2), 1/s(3), s(4))

estimating these parameters:

Amplitude =   0.503
Frequency =   0.960
Phase     =  -0.803
Offset    =  21.477

and produces this plot:

It was not even necessary to change my original code.

[Giru Mishra]

I am so sorry.. I sent a wrong data file: this one is giving error.. its giving NaN values.. and only data is plotting not the fit. Apologies

[Star Strider]

No worries.

To use your new data, in place of the original x and y assignments, substitute:

Lidx = all(isfinite(data),2);                                   % All Columns In Every Row Must Be Finite (Not ‘±Inf’ Or ‘NaN’)
x = data(Lidx,1);
y = data(Lidx,2);

to include only those rows with finite data. (This is a standard MATLAB convention, since ±Inf and NaN values will produce non-finite results.)

My code then produces these parameters:

Amplitude =   0.291
Frequency =   0.991
Phase     =  -0.702
Offset    =   3.900

and this plot:

I also edited some of the code in my earlier Comment to make it look neater, so all the comments begin in the same column. The code itself did not change.

[Giru Mishra]

woww.. you are great.. Thanks a lot.. 1 more thing.. is it possible to increase amplitude of the peak coming around 1 and also shift it to 0.8?? and also decrease the width of the central peak?

Thanks a lot again

[Star Strider]

As always, my pleasure. Thank you.

You would need to re-write the ‘fit’ function, perhaps as some version of sin(x)/x, or use the current code with a weighting vector (to give more ‘weight’ to some data) to get different results. It would depend on what your data represent, and the process that created them. A mathematical model that describes the process that created your data is always the best regression (here ‘fit’) function.

[Giru Mishra]

OH okay.. thanks..

[Star Strider]

As always, my pleasure.