EMG-CNN

Make sure to download Stockwell Transform for MATLAB (st.m)
Import your time-series EMG signal to the MATLAB environment with each fatigue level its own vector/matrix (NOT table)
Run dsp.m, it will output folders of each fatigue level containing individual windows
Ensure 'high_dir' and 'low_dir' in model.py point to these folders
Run model.py to split data, train the model, and evaluate the model

I've gotten the model to be pretty accurate even with a ridiculously small window (currently 40 samples, equivalent to 1/100th of a second of data)
After exponential drops around 10 epochs the model seems to linearly improve until asymptoting (and likely overfitting) around 200 epochs

Name		Name	Last commit message	Last commit date
Latest commit History 29 Commits
1D_CNN_models		1D_CNN_models
2D_CNN_models		2D_CNN_models
Data_raw		Data_raw
Pain_project_work		Pain_project_work
datasets		datasets
fashion_mnist_models		fashion_mnist_models
important_codes		important_codes
.gitignore		.gitignore
combining_pain_data.py		combining_pain_data.py
dsp.m		dsp.m
model.py		model.py
pain_2D_3layer.ipynb		pain_2D_3layer.ipynb
pain_2D_4layer.ipynb		pain_2D_4layer.ipynb
readme.md		readme.md
recursive_bin_read.py		recursive_bin_read.py

Provide feedback