I need help with this (seemingly simple) task.
Mainly I need identify distances between peaks of some measured signal and I know a priori that its behavior must be sin/cos function, but on the practice I have some distored signal (shown in the attached file) and as a result there are others not relevant peaks. So, the fit to the sin is good solution for me.
Relevant reference/link/MATLAB code/Python code are welcome. Thank you in advance!
Hi.
If your problem is noise reduction and you know what the frequency of sine wave is desired. you can simply filter the noise in frequency-domain with applying fft() matlab function.
Here is the main algorithm:
1- Take fft() function of your signal.
2- Remove the unwelcome artifacts or noises.
3- use inverse fft function ifft() to recover your signal in time-domain.
Hi.
If your problem is noise reduction and you know what the frequency of sine wave is desired. you can simply filter the noise in frequency-domain with applying fft() matlab function.
Here is the main algorithm:
1- Take fft() function of your signal.
2- Remove the unwelcome artifacts or noises.
3- use inverse fft function ifft() to recover your signal in time-domain.
Hi
They are few way to solve this problem but first try to remove noice using smothing procedure. The fitting may be use by some ways:
1. The simplest way is use a standard matlab function: Curve fitting to a sinusoidal function in MatlabCentral:
http://www.mathworks.com/matlabcentral/answers/121579-curve-fitting-to-a-sinusoidal-function
or
http://www.mathworks.com/matlabcentral/answers/36999-how-do-i-regression-fit-a-sinwave-to-a-dataset
2. Define a sine function as regrerssive procedure :
http://www.mathworks.com/matlabcentral/fileexchange/41246-sine-function-fit
3. Use matlab free toolbox. My favourite is FracLab 2.1 and use nonlinear fit procedure
If you have the matlab optimization toolbox available, you can use the following code:
%Create full sine-wave function for fit. 'b' is a vector with (in order)
%Amplitude (in units of signal), frequency (in Hz) and phase (in units of t) of
%the signal.
sineFit = @(b,t,Signal) b(1)*sin(2*pi*b(2)*(t+b(3)))-Signal;
%Guess starting points for signal (this is crucial, guess wrong and the fit will
%fail)...
A0 = 10; %...amplitude.
f0 = 0.5; %...fequency.
w0 = 0.7; %...phase.
%Set additional upper and lower bounds on fit parameters. This improves your fit
%in terms of accuracy and run-time. For example, let's assume the signal is a
%positive sine (Amplitude > 0 ) with a frequency between [0> and
%= 0, w
Besides the aforementioned approaches I would suggest you the following:
Tha prevalent frequency (F) of your signal must be the correspong sin/cos frequency you want (using fft and max. Once you find F then it is straight forward to find the distance between to peaks...
I believe this approach is more accurate since you do not process your input data prior to finding the distance between peaks (do not add any signal deformation)
Best, Panos
Agreed with FFT processing. If you have single sine/cosine you need only the fundamental frequency as stated by Panagiotis. In case of more complex system just filter the signal as Mohsen suggests and find the maximas from the smooth output. A very simple example in Matlab attached...
- Badges
- Science topic
- Similar topics
- Computer Science and Engineering
- Programming Languages
- Python