How can I calculate Kurtosis of Sound data?

I have sound data of people who have asthma.I made separate plots windows thanks to the start and end indexes of wheezing and non-wheezing sounds. How can I calculate the kurtosis value of each wheeze or nonwheeze windows.

wheezing starting indexes

2239                                              
19576
30338
46537
56438
69869
97067
107264
123725
134174

wheezing ending indexes

3239
24791
31829
49212
57662
72755
99823
108708
126971
136379

Answer 1

[Star Strider]

Not certain what you want.

Try this:

startidx = [2239                                              
            19576
            30338
            46537
            56438
            69869
            97067
            107264
            123725
            134174];

endidx = [3239
          24791
          31829
          49212
          57662
          72755
          99823
          108708
          126971
          136379];

framelen = endidx - startidx;
fl_mean = mean(framelen);
fl_varv = var(framelen);
fl_kurt = kurtosis(framelen);

.

[Serhat Sayim]

Let me explain clearly. For example I have a wheeze plot window between 2239 and 3239. How can I calculate the kurtosis value between these values and for this window?

[Star Strider]

I thought you wnated the statistics for the frame durations. I am still not certain what result you want.

Two functions worth exploring are kurtogram and pkurtosis in the Signal Processing Toolbox, both introduced in R2018a. See if those do what you want. If they do not, please describe what you want to do, and what result you want, in more detail.

[Serhat Sayim]

I need to find a kurtosis value of wheezing plots. First wheeze is between 2239-3239 indices and its plot is above. Please help me

[dpb]

That's exactly the code I gave you below for in-between the two indices -- you have to do a minimal amount of adjustment of those two arrays to get the start-stop for the other segments from those values.

Have you even tried it?

[Serhat Sayim]

I tried it but didnt work. Arrayfun code fails. What should I write exactly?

Kin=arrayfun(@(2239,3239)kurtosis(X(2239:3239)),iStart,iEnd);

I wrote it like that

[Star Strider]

Using a for loop:

signal = rand(1, 1.4E+5);                       % Create Signal

startidx = [2239                                              
            19576
            30338
            46537
            56438
            69869
            97067
            107264
            123725
            134174];

endidx = [3239
          24791
          31829
          49212
          57662
          72755
          99823
          108708
          126971
          136379];

for k = 1:numel(startidx)
    ktsis(k) = kurtosis(signal(startidx(k):endidx(k)));
end

Use your own vector for ‘signal’. (I created the vector here dimply to test my code.)

.

[Serhat Sayim]

It worked. Thank you so much. I've got another question about this. These were the wheezing indexes. I knew this from data structure but I have only two information about non wheeze indexes. First; its starting index, second its ending index.For example first non wheezing index starts with '3239' and ends with '19576' I do not know the indexes between them. How can I calculate the kurtosis' of non wheezing parts? Again Thank you for your help I am grateful.

[Star Strider]

As always, my pleasure!

Try this for both of them:

for k = 1:numel(startidx)
    ktsis_wheeze(k) = kurtosis(signal(startidx(k):endidx(k)));
end

for k = 1:numel(startidx)-1
    idxrng(k,:) = [endidx(k) startidx(k+1)];                                % Information To Show Indices (Delete)
    ktsis_nonwheeze(k) = kurtosis(signal(endidx(k):startidx(k+1)));
end

The first loop is the same as previously, I just re-named the variable.

See if the second loop does what you want.

(Note that ‘ktsis_nonwheeze’ it is 1 element shorter than the ‘ktsis_wheeze’ vector because I do not know how your data are organised, so I do not know whether to include any other indices.)

[dpb]

Kin=arrayfun(@(2239,3239)kurtosis(X(2239:3239)),iStart,iEnd);

That's not the code I gave you...just plug your variables in where I told you:

Kin=arrayfun(@(i1,i2)kurtosis(X(i1:i2)),iStart,iEnd);

Use whatever is your array/variable name for the data in place of "X". You plotted the data so you have the variable already, use it again.

Answer 2

[dpb]

Kin=arrayfun(@(i1,i2)kurtosis(X(i1:i2)),iStart,iEnd);

where X are your data and the two indexing arrays above for the "in group" sections.

The "out group" indices can be derived from the above in group values by adjustment of point numbers by one and adding the first, last points of the X vector. The same logic/functional form will then work for those sections.

Will leave as "exercise for Student" to consider the details...

[Serhat Sayim]

Let me explain clearly. For example I have a wheeze plot window between 2239 and 3239. How can I calculate the kurtosis value between these values and for this window?

[dpb]

That's exactly the code I gave you for in-between the two indices -- you have to do a minimal amount of adjustment of those two arrays to get the start-stop for the other segments from those values.