Dear all, how to select the values for k and f, of savitzky-golay filter?

Dear all, how to select the values for k and f, of savitzky-golay filter?

Answer 1

[Image Analyst]

You pick them to give the signal you want. There is no one right value for all situations. It's a judgment call. A longer frame width will take more points into consideration when fitting the polynomial and give a smoother curve. A higher polynomial will more accurately "hug" the actual/noisy training data, while a lower polynomial order will give a smoother, less "huggy" fit.

[Ayisha Nayyar]

Thanks a lot, but still is there any range specified for window size and polynomial...I guess I can go till 10th degree polynomial?how much window size?

[Image Analyst]

You're not going to get much smoothing if you do beyond about 3 or 4 in polynomial order. Indeed, you might actually get overfitting and false, spurious values.

The frame length must be at least 1 plus the polynomial order, and no more than the number of data points in your image. Why don't you just do a double for loop and try all combinations and see which looks best?

[Ayisha Nayyar]

Thanks again, i am going to try. Actually, i was thinking that there must be a table representing some range values of appropriate window size and corresponding polynomial.

[Image Analyst]

Try this:

clc;    % Clear the command window.
clearvars;
close all;  % Close all figures (except those of imtool.)
workspace;  % Make sure the workspace panel is showing.
format long g;
format compact;
fontSize = 20;

% Make a noisy sine wave signal
x = 1 : 3000;
period = 500
y = sin(2*pi*x/period);
noiseAmplitude = 0.8;
y = y + noiseAmplitude * rand(size(y));
subplot(2,1,1);
plot(x, y, 'b-', 'LineWidth', 2);
grid on;
title('Noisy Signal', 'FontSize', fontSize);
% Enlarge figure to full screen.
set(gcf, 'Units', 'Normalized', 'OuterPosition', [0 0 1 1]);
% Give a name to the title bar.
set(gcf, 'Name', 'Demo by ImageAnalyst', 'NumberTitle', 'Off')

% Try orders of 1-4, and frame lengths of 3 to about a quarter of your signal length.

frameLengths = 3 : 8 : 201
for polynomialOrder = 2 : 4
  for k = 1 : length(frameLengths)
    % Make sure windowWidth is long enough given the polynomial order we are using.
    windowWidth = max([polynomialOrder + 1, frameLengths(k)]);
    % Make sure windowWidth is odd
    if rem(windowWidth, 2) == 0
      windowWidth = windowWidth + 1;
    end
    caption = sprintf('Smoothed Signal with Polynomial Order = %d, windowWidth = %d', polynomialOrder, windowWidth);
    fprintf('%s\n', caption);
    
    % Now smooth with a Savitzky-Golay sliding polynomial filter
    smoothY = sgolayfilt(y, polynomialOrder, windowWidth);
    subplot(2,1,2);
    plot(x, smoothY, 'b-', 'LineWidth', 2);
    grid on;
    title(caption, 'FontSize', fontSize);
    
    promptMessage = sprintf('Do you want to Continue processing,\nor Quit processing?');
    titleBarCaption = 'Continue?';
    buttonText = questdlg(promptMessage, titleBarCaption, 'Continue', 'Quit', 'Continue');
    if contains(buttonText, 'Quit')
      break;
    end
  end
end

Adapt as needed for your signal.

[Ayisha Nayyar]

really thankful....I am trying to implement this discussion on my problem.