Filling in gaps incrementally

I have a set of time points (first column) and associated coordinates (second column) as following (for illustrative purposes):

data = [1 2 3 4 5 6 7 8 9 10; 263 264 0 0 0 0 0 0 0 731]'

A value of 0 in the coordinate column means that my tracker could not accurately capture the location of my marker. Currently, I replace the 0s with the previous value as follows (with column 1 for time, 2-3-4 are X-Y-Z coordinates and column five = -1 if tracker could not locate marker):

sz = size(data(:,1),1);
for i = 1:sz
    if data(i,5) == -1
        data(i,2) = data(i-1,2);
        data(i,3) = data(i-1,2);
        data(i,4) = data(i-1,4);
    end
end

Using the example above, I would end up with

data = [1 2 3 4 5 6 7 8 9 10; 263 264 264 264 264 264 264 264 264 731]'

However, this means that when I calculate the euclidian distance ( using sqrt((x(i)-x(i-1))^2+(y(i)-y(i-1))^2+(z(i)-z(i-1))^2) ), I end up with artificially elevated distances (using the example above) between the 9th and 10th measurement because I jump from 264 to 731.

What I would like to do is replace the 0s such that 264 --> 731 is reached uniformly. Using the example above, I want to calculate the difference between 10th and 2nd value, divide this by the number of cells I need to fill and add this number to (i-1). I've explained terribly, so using the example above:

   (731-264)/7 = 66.7
      data = [1 2 3 4 5 6 7 8 9 10; 263 264 330.7 397.4 464.1 530.8 597.5 664.3 731]'

My dataset is 80000 rows long and the 0s can occur randomly (except at the beginning) for various lengths (could have 3 zeros in a row, could also have 50). Any idea how I could code this? Any advice would be greatly appreciated.

Answer 1

[Honglei Chen]

Do you always have a fully populated first column as the time vector? If so, you can try the following to do the linear interpolation for you

nidx = data(:,2)==0;
data(nidx,2) = interp1(data(~nidx,1),data(~nidx,2),data(nidx,1),'linear','extrap')

HTH

[Nicolas Darmanthe]

I'm doing some testing but I'm really liking your approach - very straight forward! Thank you!