Fitting sine curve to data points

[Poulomi]

Hi,

I have a 343x1 discrete data matrix, which I need to fit into sine curve function. I don't know how to compute amplitude and phase of the data points. I saw some function fft but that gives a huge result 70. Can anybody please give me proper suggestion, how to fit sine curve of the form A*Sin(2*pi*f*t) or A*Sin(2*pi*f*t)+B form. Kindly help in this regard. Thanks in advance.

[TideMan]

On Sunday, January 19, 2014 12:27:08 PM UTC+13, Poulomi wrote:
> Hi,
>
>
>
> I have a 343x1 discrete data matrix, which I need to fit into sine curve function. I don't know how to compute amplitude and phase of the data points. I saw some function fft but that gives a huge result 70. Can anybody please give me proper suggestion, how to fit sine curve of the form A*Sin(2*pi*f*t) or A*Sin(2*pi*f*t)+B form. Kindly help in this regard. Thanks in advance.

Do you know f a priori?
If so, it's easy using the \ least squares facility.
If not, it's more tricky.

[Poulomi]

TideMan <mul...@gmail.com> wrote in message <f5ecd74f-2ef3-4a94...@googlegroups.com>...

No, I don't know f. Pls. explain how to do so & fitting other parameters A,..C using least squares. Look forward..

[Nasser M. Abbasi]

On 1/18/2014 10:46 PM, Poulomi wrote:

> No, I don't know f. Pls. explain how to do so & fitting other parameters
>A,..C using least squares. Look forward..
>

http://www.mathworks.com/matlabcentral/fileexchange/41246-sine-function-fit
http://www.mathworks.com/matlabcentral/answers/18446-sine-curve-fitting-from-given-dataset-using-matlab
http://www.mathworks.com/matlabcentral/newsreader/view_thread/325956
http://stackoverflow.com/questions/16110185/how-to-fit-a-curve-to-a-damped-sine-wave-in-matlab
https://groups.google.com/forum/#!topic/comp.soft-sys.matlab/hLxnvUXE_RI
http://math.stackexchange.com/questions/36725/how-to-fit-a-curve-to-a-sinusoidal-wave

[Ram Dwivedula]

Try following function:

function [freq,amp,phase]=obtain_fft(x,w1,w2,ts,varargin)
%
%gives frequency and amplitude content of the signal in vector x
%
% Usage : [freq,amplitude,phase]=obtain_fft(x,w1,w2,ts,p,'units','signal')
%
% x : signal whose fft is sought
% [w1,w2] : is frequency range of interest
% ts : sampling period in seconds
% freq : frequency in rad/sec
% amp : amplitude
% phase : phase in degrees
% p : if equal to 1 also draws amplitude plot (closes all open plots)
% if equal to 2 draws phase plot also
% 'units' : units of the elements of vector x (lbf, N, rad/sec, etc)
% 'signal' : indicates the physical quantity (tension, velocity, etc)
% p, 'units', 'signal' are optional and may be omitted

N=length(x);
dd_fft=abs(fft(x));
phase=angle(fft(x));
dd_fft=dd_fft/N*2;
NN=floor([(w1*N*ts+1):(w2*N*ts+1)]);
freq=(NN-1)/N/ts;
amp=dd_fft(NN);
phase=phase(NN)*180/pi/2;
%[w1,a1,w2,a2]=pick12natfreq(freq,amp)
if nargin==5 | nargin==6 | nargin==7% only p is specified
p=cell2mat(varargin(1));
if p==1 | p==2
if nargin==6 | nargin==7
units=char(varargin(2));
yunits=sprintf('Amplitude (%s)',units);
else
yunits='Amplitude';
end
% close all
% figure(1)
if p==2
subplot(2,1,1)
end
if nargin==7
titlestr=sprintf('of %s',char(varargin(3)));
else
titlestr='';
end
titlestr=sprintf('Frequency Content %s',titlestr);
plot(freq,amp,'r-')
ylabel(yunits)
xlabel('\omega (Hz)')
grid on
title(titlestr)
if p==2
subplot(2,1,2)
plot(freq,phase,'r-')
ylabel('Phase (degrees)')
xlabel('\omega (Hz)')
grid on
end
end
end
return




"Poulomi" wrote in message <lbf2kc$rn6$1...@newscl01ah.mathworks.com>...

[Charlie Roberts]

>
>"Poulomi" wrote in message <lbf2kc$rn6$1...@newscl01ah.mathworks.com>...
>> Hi,
>>
>> I have a 343x1 discrete data matrix, which I need to fit into sine curve function. I don't know how to compute amplitude and phase of the data points. I saw some function fft but that gives a huge result 70. Can anybody please give me proper suggestion, how to fit sine curve of the form A*Sin(2*pi*f*t) or A*Sin(2*pi*f*t)+B form. Kindly help in this regard. Thanks in advance.

If all you want to do is fit a sine, fminsearch will work very well.
I use it to for just this purpose and find it quite stable.

You will have to write a very short function that takes your
data as input and returns the residual sum of squares -- which
fminseach will seek to minimize.

The documentation for fminsearch provideds a detailed
example, function handles and all.

You will have to check for negative amplitudes and shift
the phase by pi in that case. Also, check for phases values
larger than 2*pi.

Charlie