- Used Dataset : https://github.com/nhoyh/HR_IMU_falldetection_dataset
- From the dataset we mainly included the 10 parameters which are ax, ay, az, w, x, y, z, droll, dpitch, dyaw.
Trained models :
- knn_model.pkl (binary classification)
- knn_model_all.pkl (multiclass classification)
We have evaluated with different machine learning techniques like catboost,xgboost,KNN and random forest such that we got a good accuracy in KNN which is around 98% using binary classification for fall detection and 92% using multiclass classification for activity recognition.Tested the sampled data by manually adding the data inputs into a platform of frontend(Reactjs) and backend(Flask).And also built a cross-platform mobile app (Android/iOS) using React Native with real-time fall alerts, SOS functionality, live location tracking, and emergency contact calling.