How do I randomly extract 20 segments of data from a signal?

I have a matrix of data 60001x4. Rows are data points, columns are the 4 channels recording the data. How would I extract 20 random samples of 1000 data points from each channel? Goal would be to have a 20x1 cell array where each cell is a 1000x4 matrix representing 1 1000-sample segment across the 4 channels. I have done this so far but there must be a better way to do this to make it truly random segments. I simply divided the duration of my signal by 1000 to get how many (num_segments) I would end up with. Then i used randperm to generate 20 random numbers that would correspond to each segment of the 20 out of 60 segments, in this case, that I want.

random_indices = randperm(num_segments,20);

Answer 1

[Star Strider]

If I understand correctrly what you want to do, this could work:

signal = randn(60001,4)

signal = 60001×4

-0.2521 -1.3696 -3.9248 -1.4181 -0.3019 -1.7351 -0.4578 -1.4187 2.4288 0.3358 -0.2064 0.4342 -1.0420 0.6007 0.0606 0.4438 -0.3845 -0.2802 0.5768 -0.7815 0.4380 -1.7393 -1.0484 -1.6826 2.4050 -0.1283 -2.0273 0.8010 1.2604 1.2681 1.6034 -1.2695 -1.8335 0.8577 -0.0428 0.1045 -0.3216 -0.7440 -1.1730 -1.2357

random_indices = randperm(size(signal,1)-1000, 20)

random_indices = 1×20

46774 32626 2192 55305 11347 31821 18564 46263 17113 36835 11881 54189 27679 5805 58827 48480 37787 40739 8266 57593

rowidx = [random_indices(:) random_indices(:)+999]

rowidx = 20×2

46774 47773 32626 33625 2192 3191 55305 56304 11347 12346 31821 32820 18564 19563 46263 47262 17113 18112 36835 37834

for k = 1:numel(random_indices)
    RandomSamples{k,:} = signal(rowidx(k,1):rowidx(k,2),:);
end
RandomSamples

RandomSamples = 20×1 cell array
    {1000×4 double}
    {1000×4 double}
    {1000×4 double}
    {1000×4 double}
    {1000×4 double}
    {1000×4 double}
    {1000×4 double}
    {1000×4 double}
    {1000×4 double}
    {1000×4 double}
    {1000×4 double}
    {1000×4 double}
    {1000×4 double}
    {1000×4 double}
    {1000×4 double}
    {1000×4 double}

Some of the samples would necessarilly overlap.

.

Answer 2

[David Hill]

You could just keep matrix form with a 3rd dimension

for k=1:20
  newMatrix(:,:,k)=yourMatrix(randperm(60001,1000),:);
end

[Anas Khan]

I believe this only extracts 1000 random numbers ranging from 1 to 60001. I need them to be continuous. Thanks for your input!