Assigning features to the matrix based on windowing

Hello,

At the beginning I have at my disposal file with 3 columns (signal sample, condition, device):

-297.367190000000	8	1
-295.132890000000	8	1
-282	8	1
-268.007000000000	8	1
-262.109380000000	8	1
-263.296880000000	8	1
-263.562500000000	8	1
-258.973210000000	8	1
-255.209820000000	8	1
-252.130060000000	8	1
-247.141450000000	8	1
-244	8	1
-243.025000000000	8	1
-237.704170000000	8	1
-230.770830000000	8	1
-230.597220000000	8	1
-239.697970000000	8	1
-254.455570000000	8	1
-268.051250000000	8	1
-269.360000000000	8	1
-255.611150000000	8	1

100.781000000000	3	2
54.1178190000000	3	2
25.2507100000000	3	2
23.8387100000000	3	2
46.2026210000000	3	2
49.5785290000000	3	2
20.7356310000000	3	2
-16	3	2
-73.7242630000000	3	2

Then using a the presented code for windowing I extract the following features:

fs = 256;
sbin=4; 
%sbin = 1; %1s
window=fs*sbin; 
overlap=fs;
Nwin = floor((length(data)-window)/overlap)+1;
for k=1:Nwin
    feature{1,1}(k) = var(data(1,(k-1)*overlap+1:(k-1)*overlap+window));
    [Higuchi_FD(k)] = feature2(data(1,(k-1)*overlap+1:(k-1)*overlap+window),window);   
    feature{2,1}(k) = feature2(k);
    [Katz_FD(k)] = feature3(data(1,(k-1)*overlap+1:(k-1)*overlap+window));   
    feature{3,1}(k) = feature3(k);    

for instance it gives the following results (the information is included in the rows):

    {[ 0.01 0.01 0.01 0.01 0.01 0.01 0.01 0.01 0.01 0.02 0.02 0.01 0.01 0.01 0.01 0.01 0.01 … ]}
    {[ 1.21 1.21 1.20 1.21 1.22 1.21 1.21 1.21 1.21 1.20 1.21 1.21 1.20 1.21 1.21 1.20 1.19 … ]}
    {[ 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00 … ]}

At the end I need to assign the information about condition and device to the windowed features. There 9 condition and 3 devices. How to perform that?

Regards

Elzbieta

[Voss]

Please share all code and data necessary for us to run the code you've shown here. You can upload files using the paperclip button.

Answer 1

[Shishir Reddy]

Hey Elzbieta

Assigning the information about condition and device to the windowed features can be done by determining the most frequent condition and device for each window and assigning them to the features. This can be implemented as shown in the following code snippet -

for k = 1:Nwin
    start_idx = (k-1) * overlap + 1;
    end_idx = (k-1) * overlap + window;

    windowed_data = signal_samples(start_idx:end_idx);    
    windowed_conditions = conditions(start_idx:end_idx);
    windowed_devices = devices(start_idx:end_idx);
    
    features{1, 1}(k) = var(windowed_data);
    features{2, 1}(k) = feature2(windowed_data, window); % Assuming feature2 is defined
    features{3, 1}(k) = feature3(windowed_data); % Assuming feature3 is defined
    
    conditions_windowed(k) = mode(windowed_conditions); % or conditions(start_idx)
    devices_windowed(k) = mode(windowed_devices); % or devices(start_idx)
    
end

Within each window, the condition and device are determined. As conditions or devices may vary within the window, the most frequent value (‘mode’ in MATLAB) is taken.

This approach ensures that each set of features extracted from a window is accurately labelled with the appropriate condition and device.

I hope this helps.

[Elzbieta]

Thank you for your answer. However the the order of fragments containing the same condition and device may be different from that resulting from for loop. In addition, the windows are of different lengths

Answer 2

[Image Analyst]

Why not just tack on the condition and device as the last two rows in your loop?

feature{4}(k) = data(k, 2); 
feature{5}(k) = data(k, 3);