Project Description

DeepCare: Activity Recognition via Deep Learning

The DeepCare solution allows Certified Nursing Assistants (CNAs) and nurses to monitor more closely the ambulatory status of residents at eldercare and assisted living facilities. DeepCare uses Deep Learning and other machine learning techniques to determine whether a resident is sitting, walking, standing or climbing stairs. Wearables or smartphones outfitted with the app continuously sends sensor (accelerometer, gyroscope, GPS) data to a back-end server. The system then uses the machine-learning models to accurately predict the class of activity the user is performing.

The system then continuously sends status notifications to caregivers via both a desktop web application and a mobile app. When the system detects adverse (e.g. such as falling) or unusual events (such as no motion for a long time during the day) alerts are sent both to companion apps worn by caregivers and a web application that might be installed at a nurse’s station

Registering a new resident

Activity recognition demo

Key Features

  • Continuous monitoring of residents
  • Over 95% accuracy in predicting activity state through Deep Learning techniques

Upcoming Features

  • Internal geolocation to locate resident more accurately in a facility
  • Facial recognition capability of CNA targeted mobile app to support customized care giving