Radar-APLANC: Unsupervised Radar-based Heartbeat Sensing via Augmented Pseudo-Label and Noise Contrast

This is the official code repository of our paper "Radar-APLANC: Unsupervised Radar-based Heartbeat Sensing via Augmented Pseudo-Label and Noise Contrast". The method does not require any ground truth for training.

Prerequisite

Please check requirement.txt for the required Python libraries.

Dataset and Pre-prep

Our RHB dataset maintains the same data format as the Equipleth dataset and is processed using organizer.py. If you want to use your own dataset or other datasets, please replace organizer.py.

Hierarchy of the RHB dataset

dataset
|--- RHB
|        |
|        |--- 1_1(volunteer id 1 trial 1) 
|        |         |
|        |         |--- rf.pkl(Radar data)
|        |         |--- vital_dict.npy(ground truth ppg)
|        |
|        |
|        |--- 1_2(volunteer id 1 trial 2) 
|        |
|        |
|        |
|        |--- 1_3(volunteer id 1 trial 3) 
|        |
|        |
|        |
|        |--- 2_1(volunteer id 2 trial 1) 
|        |
|        |
|        |
|
|
|--- RHB_demo_fold1.pkl(folds pickle file)
|--- RHB_demo_fold2.pkl(folds pickle file)
|--- RHB_demo_fold3.pkl(folds pickle file)
|--- RHB_demo_fold4.pkl(folds pickle file)

RHB_demo_fold[index].pkl files are the data partitions we use for cross validation. RHB_Pseudo_Generate.py is used to generate Augumented pseudo labels after the first stage. You need to ensure that the fold_path in RHB_Pseudo_Generate.py is the same as that in the first stage train.py. You can train each fold by changing the fold_path in the train.py file. If you want to test the model, please change the fold_path in test.py after each fold training and run it. This will generate a temp_est.json and temp_ get.json file to store the predicted and true values of the test set for this fold model. After testing all four fold data, run CrossValData_combine.py and CrossValidation.py for cross validation testing.

Execution

Training

Please make sure your dataset is processed as described above. You only need to modify the code in a few places in train.py and eval.py to start your training. After modifying the code, you can directly run

python train.py

Testing

After training, you can test the model on the test set. You can directly run

python test.py