Why is there this issue when I follow Waveform Segmentation using Deep Learning issues with Documentation?

I am trying to detect whether a wavform is an arc fault or nuissance fault.

For this documentation, https://www.mathworks.com/help/signal/ug/waveform-segmentation-using-deep-learning.html , in the download and prepare the data section, the documentation talks about a resizeData function, and I am looking for the same function because my code is erroring and I believe that is a reason why. Can someone direct to where the resizeData function is?

In the documentation, it talks about ECG signals, the attached signals are the signals that I am trying to train, and here is the code I have written thus far.

I PRIMARILY GET THE ERROR WHERE IT SAYS rawNet, AND I DON'T KNOW WHY

% Loading the data
datasetFolder = fullfile("Signal Training Folder")
sds = signalDatastore(datasetFolder, 'SignalVariableNames', ["faultSignalCondition", "signalLabels"])
data = preview(sds)

% Separate into Training Data
head(data{2})
M = signalMask(data{2});
plotsigroi(M, data{1}(1:1000))
rng default
[trainIdx,~,testIdx] = dividerand(numel(sds.Files), 0.7, 0, 0.3);
trainDs = subset(sds, trainIdx);
testDs = subset(sds, testIdx);

type getmask.m
trainDs = transform(trainDs, @getmask);
testDs = transform(testDs, @getmask);

transformedData = preview(trainDs)

plot(transformedData{2}(1:1000))

trainDs = transform(trainDs,@resizeData);
testDs = transform(testDs,@resizeData);
preview(trainDs)

tallTrainSet = tall(trainDs);
tallTestSet = tall(testDs);
trainData = gather(tallTrainSet);
trainData(1,:)

testData = gather(tallTestSet);

Raw Signals into the LSTM Network
% actionFlag = "Train networks"
% if actionFlag == "Download networks"
%     load("trainedNetworks.mat")
% end
load LSTM_layers_2.mat

% Training Options
options = trainingOptions('adam', ...
    'MaxEpochs',10, ...
    'MiniBatchSize',50, ...
    'InitialLearnRate',0.01, ...
    'LearnRateDropPeriod',3, ...
    'LearnRateSchedule','piecewise', ...
    'GradientThreshold',1, ...
    'Plots','training-progress',...
    'shuffle','every-epoch',...
    'Verbose',0,...
    'DispatchInBackground',true);

% Train LSTM Network
trainDataInput = trainData(:, 1)
% trainDataInput = trainDataInput{:, 1}
% trainDataInput_Curr = trainDataInput(:, 1);
% trainDataInput_RSSI = trainDataInput(:, 2);

trainDataOutput = trainData(:, 2);
trainDataOutput = trainDataOutput{1, 1}

% rawNet = trainNetwork(, trainDataOutput, layers_2, options);

Get Mask Function
function outputCell = getmask(inputCell)
%GETMASK Convert region labels to a mask of labels of size equal to the
%size of the input ECG signal.
%
%   inputCell is a two-element cell array containing an ECG signal vector
%   and a table of region labels. 
%
%   outputCell is a two-element cell array containing the ECG signal vector
%   and a categorical label vector mask of the same length as the signal. 

sig = inputCell{1};
roiTable = inputCell{2};
L = length(sig);
M = signalMask(roiTable);

% Get categorical mask and give priority to QRS regions when there is overlap
mask = catmask(M,L,'OverlapAction','prioritizeByList','PriorityList',[1 2]); %I CHANGED THIS, COULD THIS BE THE ISSUE

% Set missing values to "n/a"
mask(ismissing(mask)) = "n/a";

outputCell = {sig,mask};
end

And these are my layers, I couldn't upload any attachments, but these layers are the layers that I used

layers = [ ...
    sequenceInputLayer(28271) %in a video about this documentation, they used 1, but I did not, is this the issue?
    lstmLayer(138,'OutputMode','sequence')
    fullyConnectedLayer(4)
    softmaxLayer
    classificationLayer];

Is there anyone that can help with the resize or general issue? I don't mind discussing this issue in more detail.

Answer 1

[YANHUI LI]

function outputCell = resizeData(inputCell)
%RESIZEDATA Break input ECG signal and label mask into segments of length
%5000.
%
%   inputCell is a two-element cell array containing an ECG signal and a
%   label mask.
%
%   outputCell is a two-column cell array containing as many 5000-long
%   signal segments and label masks that were possible to generate from the
%   input data.

% Copyright 2019 The MathWorks, Inc.

    targetLength = 5000;
    sig = inputCell{1};
    mask = inputCell{2};
    
    % Get number of chunks
    numChunks = floor(size(sig,1)/targetLength);
    
    % Truncate signal and mask to integer number of chunks
    sig = sig(1:numChunks*targetLength);
    mask = mask(1:numChunks*targetLength);
    
    % Create a cell array containing signal chunks
    sigOut = reshape(sig,targetLength,numChunks)';
    sigOut = num2cell(sigOut,2);
    
    % Create a cell array containing mask chunks
    lblOut = reshape(mask,targetLength,numChunks)';
    lblOut = num2cell(lblOut,2);
    
    % Output a two-column cell array with all chunks
    outputCell = [sigOut, lblOut];

end