LoRa-HAR is a Python project for evaluating Human Activity Recognition (HAR) from LoRa radio measurements. The project was developed as part of a thesis and contains scripts for parsing LoRa measurement logs, extracting RSSI-based features, training classical machine-learning models, evaluating predictions, and generating plots and result files.
The repository includes both the source code and measurement data so that the experiments can be reproduced or tested on the provided datasets.
LoRa-HAR/
├── har_lora/
│ ├── config/ # Evaluation profiles and experiment configuration
│ ├── evaluation/ # Metrics, segment analysis, transition analysis
│ ├── features/ # RSSI feature extraction
│ ├── io/ # WiMOD / LoRa log loading utilities
│ ├── models/ # Classical ML models and baselines
│ ├── postprocessing/ # Prediction smoothing and post-processing
│ ├── preprocessing/ # Window preparation
│ ├── protocol/ # Activity schedule and label assignment
│ └── reporting/ # Plotting and export utilities
├── measurement data/ # Measurement data used for evaluations
├── out/ # Generated outputs, rankings, plots, and reports
├── run_pipeline.py # Main evaluation pipeline
├── plot_rssi.py # RSSI visualization script
├── sweep_binary_configs.py # Binary-classification configuration sweep
├── sweep_multiclass_configs.py # Multiclass configuration sweep
└── requirements.txt # Python dependencies