Help with breaking into windows based on a condition??

Hi, I have a array of numbers (0,1) which is attached in the CSV file. I want to break them into windows where the condition meets(1) for greater than or equal to 30 seconds.

If it is for only 10 seconds continuous it shouldn't be considered a window. I also want the locations for the start and stop of the windows.

Answer 1

[Image Analyst]

Assuming the spacing between elements if 1 second, you can use regionprops() to measure the starting and stopping indexes of each region:

% Read data file.
data = fileread('nte.csv');
% Skip 'condition' and convert from characters to numbers.
data = sscanf(data(10:end), '%d') 
% Get rid of regions fewer than 30 elements
data = bwareaopen(data, 30);
% Assign an ID number to each contiguous region of 1's.
[labeledRegions, numRegions] = bwlabel(data);
% Find starting and stopping indexes, plus area, for each region.
props = regionprops(labeledRegions, 'PixelList', 'Area');
% props is a structure array.
% Get all the areas into single arrays for convenience.
allAreas = [props.Area]
for k = 1 : numRegions
  starts(k) = props(k).PixelList(1, 2);
  stops(k) = props(k).PixelList(end, 2);
end

Answer 2

[John BG]

Hi Satya

convolution is very useful to answer your question

1. capturing data and display

A=importdata('NTE.csv')
L=A.data';
figure(1);stem(L);

.

2. generating window of 30 samples

mask=ones(1,30)

3. convolution

D=conv(L,mask);

4. spot windows of length 30 consecutive ones.

d_start= find(D==30)-30;
L2=zeros(1,length(L))
  L2(d_start)=1

RESULT:

L2 contains the positions of the signal where a window of 30 or more ones start.

A burst of for instance 5 ones in L2 means there is a window of 35 samples long.

Windows below 30 consecutive samples are not marked

5. Display

    figure(2);stem(L);
    hold all;
    figure(2);stem(L2,'r');

.

Satya

if you find these lines useful would you please mark my answer as Accepted Answer?

To any other reader, if you find this answer of any help please click on the thumbs-up vote link,

thanks in advance for time and attention

John BG

<mailto:[e-mail] [e-mail]>

Answer 3

[Walter Roberson]

data = csvread('nte.csv');

   Fs = 1;    %sampling rate in hz
   Samp30 = floor(Fs * 30);   %samples in 30 seconds 
   template = ones(1, Samp30) ;

   starts = strfind([0,data], [0,template]);
   ends = strfind([data, 0],[template, 0]) + Samp30 -  1;

Now starts and ends are vectors in which starts(k) gives the position of the beginning of a stretch that is long enough, and ends(k) gives the position of the last place that is set for the run.

[Image Analyst]

It's a little more complicated to get data than that because his csv file is not a standard one that csvread() can read - it throws an error. importdata() like John used, or fileread() like I used, will work with an additional line of code to throw away non-numerical data. Other than that though, this shows how to use the neat trick of using strfind() to search numerical arrays for patterns.

[Walter Roberson]

data = csvread('NTE.csv', 1) .';

works to skip the header line, since about R2014a-ish.

[Image Analyst]

That's good to know. Though it would be nice if they also allowed the 'HeaderLines' option to be consistent with all the other functions that use that method of specifying the number of lines to skip.

It's not clear that specifying the row and column to start extracting data will work now since they had that before, which worked if you had strictly numerical data, but barfed if you had any string header lines. So now it works, but those of us remembering the old way may still not know to try it again.

[Walter Roberson]

Yeh, what happened was that they quietly removed the numeric-only restriction from the documentation without a release note. If I recall correctly, string row headers are permitted now too. This still relies upon undocumented behaviour of textscan().

readtable() appears to use a different front end parser to figure out what columns mean. On the other hand readtable() documents (or at least implies) that it calls textscan(). (Well, not for .xlsx files certainly.)