This is the fall detection using machine learning
in this fall detection, I can achieve the 98% accuracy , and the precision is 96%, the recall is 98%. but it may vary when test with wearable device.
#Highlights and Specialties:
I refer the paper A Real-time Patient Monitoring Framework for Fall Detection
#dataset
use the MobiFall and MobiAct datasets, you can download it here
#file description
----client/FallDector
this is the wearable test application, it is build and running in samsung watch device. it will collect the sensor data, and send it to server
----fall_detect_svc.joblib
this is the generated model, you can load it without trainning.
----fall_detection_server.py
this is a simple server application, running it before running client application.
----tain.ipynb
this is the training process, I am lazy, and do not reorganized the code, and remove the log, you may need to prepare the machine learing environment first, and download the dataset.