/Stress-Detection-Processor

Primary LanguageJupyter Notebook

STRESS DETECTION PIPELINE WITH PHYSIOLOGICAL SIGNALS

The Stress Detection Pipeline comprises of two following steps:

  1. Bio-signal Processing and Statistical Feature Processing
  2. Train stress detection model
  3. Use the trained model to predict other data

1. Bio-signal Processing & Statistical Feature Extraction

Change the current directory to feature_extraction, which contains the necessary scripts:

$ cd ./feature_extraction

This process can be run in parallel to boost the speed of the statistical feature extraction by running the command-line separately for each physiological signal. The bio-signal processing is included in the feature extraction progress. 

$ python extract_features.py --dataset_name [DATASET_NAME] --signal [SIGNAL] --window_size [WINDOW_SIZE] --window_shift [WINDOW_SHIFT]

After all the corresponding features are extracted, running this command-line to combine the features for training:

$ python combine_features.py --dataset_name [DATASET_NAME] --window_size [WINDOW_SIZE] --window_shift [WINDOW_SHIFT]

Finally, generate the metadata which includes ground_truth and groups for subject-independent/subject-dependent training:

$ python generate_feature_metadata.py --dataset_name [DATASET_NAME] --window_size [WINDOW_SIZE] --window_shift [WINDOW_SHIFT]

If you only want to process the bio-signal and extract features for a user's data:

$ python extract_features.py --dataset_name [DATASET_NAME] --signal [SIGNAL] --window_size [WINDOW_SIZE] --window_shift [WINDOW_SHIFT] --user_id [USER_ID]
$ python generate_feature_metadata.py --dataset_name [DATASET_NAME] --window_size [WINDOW_SIZE] --window_shift [WINDOW_SHIFT] --user_id [USER_ID]

In this step, the aforementioned parameters in the command-lines are as follows:

  • DATASET_NAME: WESAD, AffectiveROAD, CognitiveDS, DCU_NVT_EXP2
  • SIGNAL: eda, bvp, temp
  • WINDOW_SIZE: 60 (by default as frequently used by many related paper)
  • WINDOW_SHIFT: 0.25 (by default as frequently used by many related paper)

2. Train Stress Detection Model

Train the stress detection model using the command-line with suitable parameters:

$ python run_classifier.py --dataset_name [DATASET_NAME] --model_name [MODEL_NAME] --detector_type [DETECTOR_TYPE] --window_size [WINDOW_SIZE] --window_shift [WINDOW_SHIFT]

In this step, the aforementioned parameters in the command-lines are as follows:

  • DATASET_NAME: WESAD, AffectiveROAD, CognitiveDS, DCU_NVT_EXP2
  • MODEL_NAME: random_forest, knn, svm, branch_neural_network
  • DETECTOR_TYPE: General (Subject-Independent Model), Personal (Subject-Dependent Model)
  • WINDOW_SIZE: 60 (by default as frequently used by many related paper)
  • WINDOW_SHIFT: 0.25 (by default as frequently used by many related paper)

If you only want to train the subject-dependent model for a target user:

$ python run_classifier.py --dataset_name [DATASET_NAME] --model_name [MODEL_NAME] --detector_type Personal --window_size [WINDOW_SIZE] --window_shift [WINDOW_SHIFT] --user_id [USER_ID]