Classify error: requires 3 arguments

Hello All I have trained an LSTM model to classify EMG signals ( one dimensional time series) to produce a class prediction. Now when trying to test the trained LSTM on a test signal , classify produces error of requiring 3 arguments. No matter how I changed the shape of the test signal nothing helped. Also predict produced error results. Could you please help.

% Training code:
% LSTM-1D classification using raw EMG signal 
% 
% Data path
path = '/home/ubuntu/Desktop/EMG data analysis/EMG signal Matlab'
parameters
numHiddenUnits = 120;
numClasses = 8;
numChannels = 1


% Now prepare training/lables dataset for LSTM training 
% Assuming sorted_emg_data is your sorted array with data and labels
% Extract the data for training
XTrain = cellfun(@(c) c.signal, sorted_emg_data(:, 1), 'UniformOutput', false);

% Extract the labels for training
TTrain = sorted_emg_data(:, 2);
% Convert the labels to a categorical array
TTrain = categorical(TTrain);

% Now XTrain contains all the EMG signals and TTrain contains the corresponding labels

% Now training the LSTM model

numHiddenUnits = 120;
numClasses = 8;
numChannels = 1


% Now define your layers with the correct number of output classes
layers = [ ...
    sequenceInputLayer(numChannels)
    bilstmLayer(numHiddenUnits, 'OutputMode', 'last')
    fullyConnectedLayer(numClasses)  % Make sure this matches the number of unique classes in TTrain
    softmaxLayer
    classificationLayer];

% Define your training options (make sure MiniBatchSize is appropriate for your dataset)
options = trainingOptions('adam', ...
    'MaxEpochs', 50, ...
    'MiniBatchSize', 3, ... % Adjust based on your hardware capabilities
    'InitialLearnRate', 0.01, ...
    'GradientThreshold', 1, ...
    'Verbose', 0, ...
    'Plots', 'training-progress');

% Train the network
net = trainNetwork(XTrain, TTrain, layers, options);



% Testing code
% 
% Loading the network 
net = load ("lstm_trained_model.mat")

% Loading data
test = load ('emg_signal_3.mat')

net.layers
length (test)
length (signal) % signal directly loaded
% Classification 
pred = classify(net, test);

[Image Analyst]

If you have any more questions, then attach your model in a .mat file and code to read it in with the paperclip icon after you read this:

TUTORIAL: How to ask a question (on Answers) and get a fast answer

[Mohamed Taha]

Got it, I updated the question and provided code. I hope that helps

Thanks

[Cris LaPierre]

Also, please share the full error message (all the red text).

[Mohamed Taha]

Hello

I am sharing the actual codes and error here and training code

The training data was (signal/labels, with signals in first column as cell array) signals are one dimension time series

[Cris LaPierre]

Can you share the result of this command?

which classify

Answer 1

[Cris LaPierre]

I cannot duplicate your error. I used this example to create a sample data set. I then trained that data using your code, and then tested it using the code in the pdfs. My conclusion is there is nothing wrong with your code. Without more details, I don't know what more we can do to help.

Here are the results I obtained when running the model on test data using the code from your pdfs.