Separating elements from a column

Hi, I am working on data which resembles like below:

B=
-0.013925234  2.83821
-0.036085514  1.61022
-0.046681829  0.97135
-0.046968672  0.89513
-0.024245858  1.81864
0.01516536  3.87455
0.041675541  5.43357
0.051087392  6.08749
0.041758625  5.71526
0.017252935  4.49994
-0.014129511  2.82299
-0.036253916  1.60009
-0.047028757  0.96068
-0.047154755  0.88619
-0.024045187  1.81487
0.015306466  3.87658
0.041772875  5.43423
0.051511911  6.10654
0.042270005  5.73995
0.017497553  4.51872
-0.013852278  2.84432
-0.035969299  1.61081
-0.046629988  0.97547
-0.046837352  0.90091
-0.024421296  1.80676
0.015151971  3.87475
0.042088386  5.44538
0.051397067  6.10449
0.042292149  5.73531
0.017802252  4.521
-0.013961625  2.82833
-0.036328589  1.60363

I have thousands of such values. The problem is that I have to look for the minimum value in the first column and starting from that going forward until I get another minimum value (as close as possible to the first one) and separate the corresponding elements in both columns. This process has to be repeated over the whole data. So if I look at the given data, this is how output will look like:

a =
-0.046968672  0.89513
-0.024245858  1.81864
0.01516536  3.87455
0.041675541  5.43357
0.051087392  6.08749
0.041758625  5.71526
0.017252935  4.49994
-0.014129511  2.82299
-0.036253916  1.60009
-0.047028757  0.96068
-0.047154755  0.88619

b =
-0.047154755  0.88619
-0.024045187  1.81487
0.015306466  3.87658
0.041772875  5.43423
0.051511911  6.10654
0.042270005  5.73995
0.017497553  4.51872
-0.013852278  2.84432
-0.035969299  1.61081
-0.046629988  0.97547
-0.046837352  0.90091

Is there any tool / method available in matlab for such a task?

Answer 1

[John BG]

Waqar

It seems you have the input and output signals of an amplifier.

The way to solve this question is to

1.- invert the first column of B

2.- find mins by spotting peaks of inverted signal with findpeaks

3.- apply the location of found peaks to the original image

One can always build a custom function but there is standard no valleys function or minsearch option in function findpeaks .

Before trying point one I found command cummin, but Cumulative minims search command cummin doesn't help that much because is returns clusters of minims rather than the minims you want

B=[-0.013925234  2.83821
-0.036085514  1.61022
-0.046681829  0.97135
-0.046968672  0.89513
-0.024245858  1.81864
0.01516536  3.87455
0.041675541  5.43357
0.051087392  6.08749
0.041758625  5.71526
0.017252935  4.49994
-0.014129511  2.82299
-0.036253916  1.60009
-0.047028757  0.96068
-0.047154755  0.88619
-0.024045187  1.81487
0.015306466  3.87658
0.041772875  5.43423
0.051511911  6.10654
0.042270005  5.73995
0.017497553  4.51872
-0.013852278  2.84432
-0.035969299  1.61081
-0.046629988  0.97547
-0.046837352  0.90091
-0.024421296  1.80676
0.015151971  3.87475
0.042088386  5.44538
0.051397067  6.10449
0.042292149  5.73531
0.017802252  4.521
-0.013961625  2.82833
-0.036328589  1.60363]

figure(1)
yyaxis left;plot(B(:,1)')
yyaxis right;plot(B(:,2)')
grid on

a =[-0.046968672  0.89513
-0.024245858  1.81864
0.01516536  3.87455
0.041675541  5.43357
0.051087392  6.08749
0.041758625  5.71526
0.017252935  4.49994
-0.014129511  2.82299
-0.036253916  1.60009
-0.047028757  0.96068
-0.047154755  0.88619]

figure(2)
yyaxis left;plot(a(:,1)')
yyaxis right;plot(a(:,2)')
grid on

d=cummin(B(:,1)')
B1=B(:,1)';B2=B(:,2)';
nd=0;for k=1:1:numel(d)  v0=find(B1==d(k));nd=[nd v0]; end;

figure(3);plot(B(:,1),'b');hold all;plot(B(:,2),'r');legend('B 1st column','B second column');grid on
hold all;figure(3);plot(nd+1,B1(nd+1),'bo')

1.- Inverting input signal

B1max=max(B1);
B1inv=B1max-B1;

2.- spotting mins

[mins locmins]=findpeaks(B1inv);

3.- Applying found mins to signal

1st cycle of input signal:
B1([locmins(1):1:locmins(2)])
 =
  Columns 1 through 3
  -0.046968672000000  -0.024245858000000   0.015165360000000
  Columns 4 through 6
   0.041675541000000   0.051087392000000   0.041758625000000
  Columns 7 through 9
   0.017252935000000  -0.014129511000000  -0.036253916000000
  Columns 10 through 11
  -0.047028757000000  -0.047154755000000

1st cycle of output signal, the one you are asking for

B2([locmins(1):1:locmins(2)])
 =
  Columns 1 through 3
   0.895130000000000   1.818640000000000   3.874550000000000
  Columns 4 through 6
   5.433570000000000   6.087490000000000   5.715260000000000
  Columns 7 through 9
   4.499940000000000   2.822990000000000   1.600090000000000
  Columns 10 through 11
   0.960680000000000   0.886190000000000

2nd cycle of the output signal:

B2([locmins(2):1:locmins(3)])
 =
  Columns 1 through 3
   0.886190000000000   1.814870000000000   3.876580000000000
  Columns 4 through 6
   5.434230000000000   6.106540000000000   5.739950000000000
  Columns 7 through 9
   4.518720000000000   2.844320000000000   1.610810000000000
  Columns 10 through 11
   0.975470000000000   0.900910000000000

.

Waqar

if you find my answer useful would you please mark it as Accepted Answer.

To any other reader, please 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]>

[Waqar Qureshi]

Excellent and thanks

Answer 2

[Image Analyst]

B is your input data. So you find the min of its column #1. Then extract from that row downwards, and find the min in that subset. Then extract another subset, and so on. OK, fine. Easy enough.

B=[...
  -0.013925234  2.83821
  -0.036085514  1.61022
  -0.046681829  0.97135
  -0.046968672  0.89513
  -0.024245858  1.81864
  0.01516536  3.87455
  0.041675541  5.43357
  0.051087392  6.08749
  0.041758625  5.71526
  0.017252935  4.49994
  -0.014129511  2.82299
  -0.036253916  1.60009
  -0.047028757  0.96068
  -0.047154755  0.88619
  -0.024045187  1.81487
  0.015306466  3.87658
  0.041772875  5.43423
  0.051511911  6.10654
  0.042270005  5.73995
  0.017497553  4.51872
  -0.013852278  2.84432
  -0.035969299  1.61081
  -0.046629988  0.97547
  -0.046837352  0.90091
  -0.024421296  1.80676
  0.015151971  3.87475
  0.042088386  5.44538
  0.051397067  6.10449
  0.042292149  5.73531
  0.017802252  4.521
  -0.013961625  2.82833
  -0.036328589  1.60363];
startingRow = 1;
[rows, columns] = size(B)
% Get the min of column 1
[minValues(1), indexesOfMins(1)] = min(B(startingRow:end));
fprintf('The first min = %f and occurs at row %d.\n', minValues(1), indexesOfMins(1));
for k = 2 : rows
  % Start searching from the last min downwards.
  startingRow = indexesOfMins(k-1)+ 1;
  % Get the min of column 1
  [minValues(k), indexesOfMins(k)] = min(B(startingRow:end));
  % Add in the offset
  indexesOfMins(k) = indexesOfMins(k) + startingRow - 1;
  fprintf('The #%d min = %f and occurs at row %d.\n', k, minValues(k), indexesOfMins(k));
  % Break if we've hit the end of the column
  if indexesOfMins(k) >= rows
    break;
  end
end
minValues
indexesOfMins
msgbox('Done!');

But what are your outputs, the badly-named "a" and "b" ???

[Waqar Qureshi]

Yes. I have to later on interpret this data. I am trying to test your solution. Thanks