How to create a moving average filter to a specific signal

Hi

I wanna apply a moving average filter to my temperature dataset, it is a 92-day daily temperature dataset. How can I create the filter?

Thanks a lot!

Answer 1

[Chunru]

x = readmatrix("2013.txt");
n = 7;   % filter order
y = filter(ones(n, 1)/n, 1, x);

plot(x); hold on;
plot(y);
legend('Original', 'MovAve')

To get rid of delay:

figure
y = filtfilt(ones(n, 1)/n, 1, x);
plot(x); hold on;
plot(y);
legend('Original', 'MovAve')

[Yian Chen]

Thanks! Is there any way to get rid of the time delay?

[Yian Chen]

And I apply [1/3 1/3 1/3] as a window function to my data, after the fourier transform of the result, I find there is a 'zeros' in my plot, what does that mean?

[Chunru]

See above for removing delay using filtfilt (doc it for more details).

Answer 2

[Image Analyst]

Look at the movmean() function.

[Yian Chen]

Thanks! I apply [1/3 1/3 1/3] as a window function to my data, after the fourier transform of the result, I find there is a 'zeros' in my plot, what does that mean? The plot is above.

[Image Analyst]

@Yian Chen I don't see any zeros, and you might want to use more than 3 elements:

y = readmatrix('2013.txt')
plot(y, 'b.-', 'LineWidth', 2, 'MarkerSize', 20);
grid on;
xlabel('Index', 'FontSize',fontSize)
% Filter it
windowWidth = 9
smoothy = movmean(y, windowWidth);
hold on;
plot(smoothy, 'r.-', 'LineWidth', 2, 'MarkerSize', 20);
legend('Original', 'Smoothed')

You could also try sgolayfilt() to smooth it. You can choose an order like 2 or 3 to follow the original data better than movmean() which uses a linear fit. An order of 2 or 3 will fit a quadratic or cubic to the data in each window position.