Splitting the signal into different parts.

I am trying to analyze and apply filter to a signal(measurements). But the problem is while recording the measurements, between 2 timestamps the value was not recorded, so the program used to record these measurements just connected these 2 loose ends, this is a problem for my algorithm. so i want to split the signal into multiple parts based on this abnormality. I need to do this automatically. Below is the image of the abnormality.

Below i have shown a pic of plot(diff(t)).

So my signal will now divided into 3 parts. Any idea how to do this will be appreciated.

[Image Analyst]

Explain why there is a second gap of 1 second along the t vector yet no such gap seems to appear when you plotted signal vs. t in the top plot. Only one time gap of 1 second seems to show up in the top plot.

Which t index, at 3500 or 7500, is the one that shows up going from t=20 to t=21 on the top plot?

[JA]

I just zoomed in to show how it looks, there is actually 2 time gaps.

[JA]

the t index around 7500, is the one showing from t= 20, to t = 21.

Answer 1

[goerk]

If you have a signal like this shown in Figure2 you can use this information.

 changeDetected = [false, diff(t)>0.1]; %get the points with a time step
 categoryID = cumsum(changeDetected); %get the category of the time-slot (starting from 0)
 nofCategorys = categoryID(end)+1; % number of found time-slots

 for ii = 1:nofCategorys
   mask = categoryID == (ii-1);  % get a mask for the current category
   extractedSignal = signal(mask); % get the signal of the current time slot
   erg(ii)=yourAlgorithm(extractedSignal); % make your calculations
 end

where your input data are the vetors 'signal' and 't'.

[JA]

Sorry i am pretty new to matlab, while excuting this

changeDetected = ['false', diff(t) > 0.7];

I am getting

Error using char
Conversion to char from logical is not possible.

Can you tell me what exactly is the syntax.

[goerk]

The problem is that you use quotes around false. false is a boolean value and 'false' is a string.

Detailed description:

 diff(t) > 0.1

The idea behind that is that you compare the discrete derivation of the time with a threshold of 0.1s. So at the points where data is missing you get a true as result the whole rest of the vector is false;

The diff command return a vector where the length is reduced by 1 in comparison with the input vector. As we want to use it as a mask for the signal we have to add an additional element to the resulting vector.

 [false diff(t)>0.1]

I am not completely sure if you have to append the element at the beginning or at the end. (for the end use: [diff(t)>0.1 false]) But you can check this with

 t(changeDetected)

this command should return the start time of the second and the third time slot.

[JA]

I did use without quotes, as boolean. still i get this error.

changeDetected = [false, diff(t)>0.1];

 Error using horzcat
CAT arguments dimensions are not consistent.

My t is a 12000x1 array

[goerk]

Then use

 changeDetected = [false; diff(t)>0.1];

Note: The comma is replaced with a semicolon.

[JA]

Thank you for your time, last question! in the code.

   for ii = 1:notCategorys

What is notCategorys

[JA]

Did you mean nofCategorys

[goerk]

Yes, you can loop through all categorys (time-slots) and analyse the specific signal part. I fixed this typo in the statement above.