Matlab IIR filter implementation

Question

I need to implement the Matlab filter(b,a,x) function in Java. Before I do this, I would like to rewrite the function in Matlab before translating into Java. So far, writing out all the coefficients works really well. See below:

x = load('data.csv');
windowSize = 10; 
b = (1/windowSize)*ones(1,windowSize); % b coefficient - a is unused 
i = 1;
for n = length(b) : length(x) % yes I know I am losing my first 10 samples
 y(i,:) = [ b(1)*x(n) + b(2)*x(n-1) + b(3)*x(n-2) + b(4)*x(n-3) + ... 
 b(5)*x(n-4) + b(6)*x(n-5) + b(7)*x(n-6) + b(8)*x(n-7) + ...
 b(9)*x(n-8) + b(10)*x(n-9) ];
 i = i+1; % used for indexing calculated results into new array
end

However, I want the loop contents to automatically update when I change the window size. In the above example, if I were to increase the windowsize by a factor of two, then I would have to write out twice as many multiples of b and x within the for loop. Which is a terrible solution to my problem. Please advise.

Answer 1

So this solution works:

x = load('data.csv');
windowSize = 10; 
b = (1/windowSize)*ones(1,windowSize);
i = 1;
for n = windowSize: length(x)
 m = 0;
 tempSum = 0;
 for j = 1 : windowSize
 temp = b(j)*x(n-m);
 tempSum = tempSum + temp;
 m = m + 1;
 end
 y(i,:) = [tempSum]
 i = i+1;
end

See results below:

windowsize10

windowsize30

Instead of having really inaccurate results in the beginning I just skip these. It's important to note that you are losing an amount of samples equal to the number of b coefficients in this solution. So make sure to rescale the X-axis.

Results were compared using Matlabs C = setdiff(y,yy) function and found no difference between the proposed solution and Matlabs filter(b,a,x) function.

• Royi: There are much better ans simpler ways to implement Box Filter. If this is all you need, you can get much better.