EMG tic classification & prediction

Code accompanying our paper for preprocessing, dataset construction, and supervised training of classification / prediction models on multichannel EMG windows.

Environment

Option A — pip:

pip install -r requirements.txt

Option B — conda: use environment.yml as a starting point and remove packages that do not match your OS (the file was exported from a development machine).

Data layout

After obtaining data (see DATA.md), place it under raw_data/ as documented there. Defaults in the Python scripts assume raw_data and write processed splits to data/.

Typical workflow

1. Preprocess (classification + prediction, multiple window lengths):

   bash scripts/generate_data.sh

2. Train classifiers / predictors:

   bash run_classify.sh
   bash run_predict.sh

   Set PYTHON_BIN if your interpreter is not python on PATH, e.g. PYTHON_BIN=python3 bash run_classify.sh.

Checkpoint paths

run_classify.sh / run_predict.sh use --skip_pretrain on the supervised pretrain stage inside train_*.py and save finetuned weights to checkpoints/.

License

TBD