Identify the duration of zero values and count the number of such zero regions

Hello,

I have a signal shown above.

1. First I want to identify the duration(length) of each zero valued regions represented as T, T1, T2, T3 etc., (You may generate any signal with zeroed region similar to my graph)
2. Count how many such zero valued regions are there in 1 sec(or within a specified time).

Any help interms of code is well appreciated. Thank you in advance.

Answer 1

[Davide Masiello]

Use regionprops

time = 0:100;
signal = rand(size(time));
signal([3:6,15:21,43:55,61:64,87:92]) = 0;
plot(time,signal)

r = regionprops(signal==0,"SubarrayIdx");
for k = 1:length(r)
    time_indexes = r(k).SubarrayIdx{2};
    durations(k) = time(time_indexes(end))-time(time_indexes(1));
end
durations

durations = 1×5

3 6 12 3 5

Obviously, the number of zero regions is the length of the array durations.

[Reji G]

Thank you.

But for my data it is not working as expected. Herewith I've attached my data table. Plz try this.

[Davide Masiello]

That's because it seems you don't have any set of consecutive zeros, but rather many isolated ones

T = readmatrix('data.csv');
time = T(:,1);
signal = T(:,2);
plot(time,signal)

find(signal==0)  % indexes of signal array where signal is zero

ans = 4795×1

4372 4402 4411 4417 4418 4420 4422 4429 4430 4432

What you could do is define a threshold to define small values as equal to zero. Be careful, depending on the threshold value you might significantly change your signal.

thr = 0.5;
signal(signal > -thr & signal < thr) = 0;
plot(time,signal)

r = regionprops(signal==0,"SubarrayIdx");
for k = 1:length(r)
    time_indexes = r(k).SubarrayIdx{1};
    durations(k) = time(time_indexes(end))-time(time_indexes(1));
end

Because of the noise of your signal, it will still find single zero values or very small zero intervals, which I assume you don't want to see.

Set a new threshold defining that you want only intervals that are larger than a certain time span.

durations(durations<0.0005) = []

durations = 1×5

0.0009 0.0011 0.0032 0.0012 0.0024

In this way, it finds only the intervals that you can see with the naked eye in the plot (in fact, it's five of them).

By fiddling around with the various thresholds, you can refine your code further.

[Image Analyst]

Or, more simply

zeroRegions = (signal == 0); % Identify elements that are exactly 0.
% Throw out runs less than, say, 3 elements long.
zeroRegions = bwareaopen(zeroRegions, 3);
% Measure what's left (regions 3 or more elements long).
r = regionprops(zeroRegions,"Area");
durations = [r.Area] % Extract all run lengths from structure into vector.