How can I vectorize my Jaro Distance code to run more efficiently?

Hello all!

I have written a code that will perform the Jaro distance between two strings A and B. My code however uses a lot of for loops and if statements. It looks like this:

 [~,m] = size(A);
 [~,n] = size(B);

 sbools = ones(1,m);
 tbools = ones(1,n);

 sprime = 'X';
 tprime = 'X';

 H = floor(min(m,n)/2);

 %A searching on B

 for i = 1:m
    for j = (i-H):(i+H)
        if(j > n)
            break
        end
        if(j>0)
            if (s(1,i) == t(1, j))
                if(tbools(1,j))
                    sprime = [sprime s(i)];
                    tbools(1,j) = 0;
                    break;
                end
            end
        end
    end
end

 %B searching on A

 for i = 1:n
    for j = (i-H):(i+H)
        if(j > m)
            break
        end
        if(j>0)
            if (t(1,i) == s(1, j))
                if(sbools(1,j))
                    tprime = [tprime t(i)];
                    sbools(1,j) = 0;
                    break;
                end
            end
        end
    end
end

 [~,mprime] = size(sprime);
[~,nprime] = size(tprime);
if((mprime > 1) && (nprime > 1))
    sprime = sprime(1,2:(mprime)); %Removes Pre-load
    tprime = tprime(1,2:(nprime)); %Removes Pre-load
else
    out = 0;
    return;
end

 [~,mprime] = size(sprime);
 [~,nprime] = size(tprime);

 if(mprime ~= nprime)
    out = 0;
    return;
    %    error('Jaro "matching character" string length does not align');
 end

 runlen = min(mprime, nprime);
 for i = 1:runlen
    if(sprime(1,i) ~= tprime(1,i))
        transp = transp + 1;
    end
end

 %transp = transp + abs(mprime-nprime);

 transp = transp/2;

 if(mprime == 0 && nprime == 0)
    out = 0;
    return
 elseif(nprime == 0)
    out = (1/3)*((mprime/m) + (nprime/n) + ((mprime - transp)/(mprime)));
 else
    out = (1/3)*((mprime/m) + (nprime/n) + ((nprime - transp)/(nprime)));
end

This is my complete code. The first loop searches for the letters in A that match a window of the letters in B. If there is a match it will save that letter from A and switch the boolean to false so the letter in B cannot be used twice. The same for the second part. My question: is there a better way to write this? I am only interested in a solution to the Jaro or Jaro-Winkler distance since I will be using both. A good description of the metric can be found at:

   http://www.vinc.be/www-pdf/10-11_SINF2356-Mahalanobis-JaroWinckler-nDollar.pdf

and

 http://en.wikipedia.org/wiki/Jaro%E2%80%93Winkler_distance

Thank you for any help in advance!

[Azzi Abdelmalek]

Give a short example with expected result

[Walter Roberson]

Is it intended to implement the original Jaro metric, or the Jaro-Winkler metric ? If it is the original Jaro metric, we are going to need a link to a formal description of the metric, as people only seem to talk about Jaro-Winkler.

[Image Analyst]

How big are these strings and how much time are you looking to shave off? (Use tic and toc to do timing.)