moving average of unevenly spaced time data

Hello,

I have the following data, and need to average the signal data as per defined window size. Any help?

signal= [0.1 0.2 0.15 0.3]; % volt reading

time_signal = [3.5 5.5 6.5 10.5];% seconds after event, the time when signal was recorded

time_regular = [ *1 2 3 4* 5 6 7 8 *9 10 11 12 13*]; time in seconds

Window_size is 4, 4 and 5. I want to average the signal data that falls within first 4 seconds, then next 4 seconds and finally in next 5 seconds. I highlighted the window sizes in time_regular array.

The answer is 0.1, 0.175 and 0.3. The calculation is done as follows. In first 4 seconds there is only one signal data point that falls within first 4 seconds, therefore first value is 0.1. In next 4 seconds, there are two signal data points (0.2 and 0.15), therefore the average of these two points is 0.175. In next 5 seconds, there is only one signal data point (0.3), therefore the average is 0.3. I can do this using for loop, but I have 100s of window sizes and big data.

Any matlab function can do this trick?

Answer 1

[Andrei Bobrov]

t = randi([3 5],6,1); % your time intervals in seconds
n = cumsum(t);
s = rand(n(end),1);
s = s.*(s < .3); % your signal

ii = zeros(n(end),1);
ii(n - t + 1) = 1;
s(s == 0) = nan;
out = accumarray(cumsum(ii),s,[],@(x)mean(x,'omitnan'));

[Research]

This is great. Thanks. On second line variable 'n' is undefined, so code needs little revision. Move line 4 statement up. Also, note that code works only if variable 't' has integer values.

[Andrei Bobrov]

Thank you Research! I am fixed.

[Andrei Bobrov]

Other varint here 't' with don't integers

t = cumsum(.6*rand(40,1) + .5); % time 
s = .3*rand(40,1) + .4;
s = s.*(s > .5); % your signal
k = randi([3 6],6,1);
tt = k/sum(k);
ii = [0;cumsum(t(end)*tt)]; % time-intervals boundaries

jj = discretize(t,ii);
s(s == 0) = nan;
out = accumarray(jj(:),s,[],@(x)mean(x,'omitnan'));

[Julie van der Hoop]

Can I ask why the k is it setting unequal time bins? Isn't the goal to have equal time bins across the series (i.e., compute over 4 seconds)?