moving(sliding) window average to create base line in Matlab - matlab

Question

I need to compute the moving window average with a window size of 20 samples.That window should move throughout the data. Example: For the first 20 samples it will calculate the average and the window moves from 0 to and 20 to 21 and calculate the average from 1 to 21 samples and further it should move on.
I have written a code which 'x' variable is loaded with .dat file and a for loop has been written to calculate mean. The code as follows
clear all;
close all;
x= load ('cpp1500.dat');
for i=1:length(x)
s(i)=sum(x(1:i));
r(i)=s(i)./i;
end
plot(x,'R')
hold on;
plot(r)
Please suggest me a method to calculate the moving window average.
Graphs of samples v/s sensor data

Answer 1

In your code you seem to have mixed up some of the indexing. To answer the question (this is not the recommended version, for that refer to the version further below)
clear all;
close all;
x= load ('cpp1500.dat');
wndSize = 20;
for i=1:length(x) - wndSize
s(i)=sum(x(i:i + wndSize));
end
r = s ./ wndSize
plot(x,'R')
hold on;
plot(r)
Matlab usually punishes looping over arrays with high runtimes. You want to perform a 1d-filtering with a box filter of size 20 where Matlab happens to have efficiently implemented functions. This should do the trick:
windowSize = 20;
b = (1/windowSize)*ones(1,windowSize);
a = 1;
r = filter( b, a, x )
Beware that the implementation of filter is slightly different regarding the handling of elements from x that are close than windowSize to the ends of the array. However, it yields a usable sliding average.
See also: https://de.mathworks.com/help/matlab/ref/filter.html?requestedDomain=www.mathworks.com

Answer 2

That's very easy to do. You would just use the index of your for loop to help you compute that for you. Instead of going from 1 to i, you would go from i to i + 19. You will need to make sure that your for loop doesn't go beyond the bounds of the signal though, so you will need to iterate up to length(x) - 19. You also need to divide your signal by 20 before you go to the next window:
clear all;
close all;
x = load ('cpp1500.dat');
for i = 1 : length(x) - 19 % Change
s(i) = sum(x(i : i + 19)); % Change
r(i) = s(i) / 20; % Change
end
plot(x,'R');
hold on;
plot(r)
However, if I can recommend something, don't use a loop here. Use filter and specify the right side coefficients as all 1s while the left coefficient is 20. You also have to be cognizant that there will be a delay because it starts immediately filtering the signal before you have collected 20 samples to get a moving average. Therefore you will need to remove the first 19 samples from the output after you're done:
clear all;
close all;
x = load ('cpp1500.dat');
r = filter(ones(20, 1), 20, x(:));
r = r(20 : end);
plot(x, 'R');
hold on;
plot(r);
Note that I have no idea what structure your loaded vector is, so I've ensured that it's a column vector.

Answer 3

If you R2016a or later, you can use the movmean function to do this.
r = movmean(x, [0 19]);
will take a 20-point moving average starting at the current point and using 19 points to the left. There are other options for the function, depending on how you want to align the window, handle endpoints, tec.