Row insertion and interpolation

Hi,

I have a dataset as follows

data = [0 135 140 500; 0.03 139 120 485; 0.06 133 127 489; 0.09 131 122 450; 0.15 160 149 503]

where column 1 is time in seconds, 2 is displacement along X, column 3 displacement along Y and column 4 displacement along Z. As you can see, I should have values every 0.03s. Sometimes however one row (or more!) is lost, and I would like to insert the missing info using a linear extrapolation for all columns.

Currently, I detect the missing rows by calculating tdiff

s = height(data);
t=data(:,1);
tdiff=zeros(s,1);
for i=2:s %start at cell 2 cell because 1 is first measurement
    tdiff(i,1)=(t(i)-t(i-1));
end

and subsequently detect the location in time (locdroppedpose) and number (nlostpose) of lost rows (PoseT is how big the interval should be, in this case 0.03)

droppedpose=tdiff(tdiff>(PoseT));
locdroppedpose=t(tdiff>(PoseT));
nlostpose=droppedpose./(PoseT);

What I would like to do is insert into data the necessary number of rows (floor(nlostpose)) in the correct location (loclostpose).

I'm quite new to matlab and am not sure what the best ay to do this is. Any advice is appreciated.

Thank you!

Answer 1

[Star Strider]

Try this:

data = [0 135 140 500; 0.03 139 120 485; 0.06 133 127 489; 0.09 131 122 450; 0.15 160 149 503];

ti = min(data(:,1)) : 0.03 : max(data(:,1));                            % Create New Time Vector
datai = [ti'  interp1(data(:,1), data(:,2:end), ti, 'linear')]          % Interpolate & Concatenate To Create New Matrix

datai =
         0  135.0000  140.0000  500.0000
    0.0300  139.0000  120.0000  485.0000
    0.0600  133.0000  127.0000  489.0000
    0.0900  131.0000  122.0000  450.0000
    0.1200  145.5000  135.5000  476.5000
    0.1500  160.0000  149.0000  503.0000

It uses the interp1 function with a time vector of the times you want to interpolate your data. (If you are doing signal processing, the Signal Processing Toolbox resample function is most appropriate.)

[Nicolas Darmanthe]

Once again, thank you for your answer. And I will certainly look into that toolbox to see if I can find anything interesting.

[Star Strider]

As always, my pleasure.

The resample function is best for signal processing applications because it incorporates an anti-aliasing filter. If you’re not doing signal processing, you likely don’t need that.