noelplacencia/Smart_Stethoscope_CCNY_SD

# Smart Stethoscope Wireless AI-assisted stethoscope for real-time respiratory and cardiac anomaly detection. Sensor data is processed at the edge on an ESP32, transmitted over BLE to a Raspberry Pi 4, and classified using a Random Forest model trained on the ICBHI 2017 and CirCor datasets. Built for EE 59866/59868 Senior Design — The City College of New York. ## How it works Sensors (MEMS mic, Piezo, ECG, SpO₂, IMU) → ESP32 (filtering, DSP, feature extraction) → BLE → Raspberry Pi 4 (Random Forest inference) → Dashboard (waveforms, anomaly flags, patient records) ## Repository structure smart-stethoscope/ ├── firmware/ # ESP32 Arduino code │ └── main/ │ └── main.ino # Sensor collection, DSP, BLE transmission ├── rpi/ # Raspberry Pi Python code │ ├── inference.py # BLE (Bluetooth) receiver + RF (Recieved Feature) model inference │ └── dashboard.py # Real-time display and patient logging ├── ml/ # Model training pipeline │ ├── data/ # Not committed — see Datasets section below │ │ ├── rf_model_lung.joblib # Trained lung RF model │ │ ├── rf_model_heart.joblib # Trained heart RF model │ │ ├── rf_model_piezo.joblib # Trained piezo RF model │ │ ├── scaler_lung.joblib # StandardScaler for lung features │ │ ├── scaler_heart.joblib # StandardScaler for heart features │ │ └── scaler_piezo.joblib # StandardScaler for piezo features │ ├── notebooks/ # Exploratory analysis │ ├── extract_features_lung.py # 38-feature extraction — ICBHI lung sounds (MEMS mic) │ ├── extract_features_heart.py # 17-feature extraction — CirCor heart sounds │ ├── extract_features_piezo.py # 38-feature extraction — ICBHI lung sounds (piezo sensor) │ ├── train_lung.py # Train and evaluate lung RF classifier │ ├── train_heart.py # Train and evaluate heart RF classifier │ └── train_piezo.py # Train and evaluate piezo RF classifier ├── docs/ # Diagrams, reports, presentations └── README.md ## Setup ### ESP32 (firmware) ### Raspberry Pi (inference) pip install bleak scikit-learn numpy joblib python rpi/inference.py Update `ESP32_ADDRESS` in `inference.py` with your device's BLE MAC address (find it by running `python -m bleak scan`). ### ML training (laptop) pip install librosa scipy scikit-learn imbalanced-learn numpy pandas matplotlib joblib # Lung sound model — MEMS mic (ICBHI 2017, 38 features) python ml/extract_features_lung.py # → ml/data/features_lung.csv python ml/train_lung.py # → ml/data/rf_model_lung.joblib, scaler_lung.joblib # Heart sound model — murmur detection (CirCor DigiScope, 17 features) python ml/extract_features_heart.py # → ml/data/features_heart.csv python ml/train_heart.py # → ml/data/rf_model_heart.joblib, scaler_heart.joblib # Piezo sensor model — chest wall vibration (ICBHI 2017, 38 features) python ml/extract_features_piezo.py # → ml/data/features_piezo.csv python ml/train_piezo.py # → ml/data/rf_model_piezo.joblib, scaler_piezo.joblib Copy the `.joblib` files to the Pi before running inference. ## Datasets Datasets are **not committed** to this repo due to file size. Download them manually and place in `ml/data/`: | Dataset | Use | Link | |---------|-----|------| | ICBHI 2017 | Lung sound classification (normal, wheeze, crackle) | [bhichallenge.med.auth.gr](https://bhichallenge.med.auth.gr) | | CirCor DigiScope | Heart murmur detection | [physionet.org/content/circor-heart-sound](https://physionet.org/content/circor-heart-sound/1.0.3/) | | MIT-BIH | ECG arrhythmia classification | [physionet.org/content/mitdb](https://physionet.org/content/mitdb/1.0.0/) | ## Hardware | Component | Role | Interface | |-----------|------|-----------| | SPH0645LM4H-B (MEMS mic) | Primary heart/lung audio | I²S | | Ambient microphone | Noise reference for subtraction | I²S | | Piezo sensor | Chest wall vibration | ADC | | AD8232 | ECG / heart timing | ADC | | MAX30102 | SpO₂ and pulse rate | I²C | | MPU-6050 | Motion artifact detection | I²C | | ESP32 | Edge DSP + BLE transmission | — | | Raspberry Pi 4 | AI inference + dashboard | — | ### Branch naming - `ml/` — model training and feature extraction - `firmware/` — ESP32 code - `rpi/` — Raspberry Pi code - `docs/` — reports, diagrams, slides ## Team | Name | Role | |------|------| | Noel Placencia | Team Lead | | Ulash Kundu Joy | Hardware Lead | | Jason Corona | Filtering Lead| | Saqlain Warrris | AI Lead| | Puran Chaudharry | Data Lead | ## License For academic use only. Not intended for clinical deployment.