Mah1ndra/FitnesssMonitor

Human Activity Recognition is an Android application that users can use to track their daily activities like walking, jogging, sitting, standing, walking upstairs and walking downstairs. The application uses a Convolutional Neural Network (CNN) to predict user activity automatically and stores the information in a database stored on the phone. The users can then choose to visualize the statistics.

Human Activity Recoginition

Human Activity Recognition is an Android application that users can use to track their daily activities like walking, jogging, sitting, standing, walking upstairs and walking downstairs. The application uses a Convolutional Neural Network (CNN) to predict user activity automatically and stores the information in a database stored on the phone. The users can then choose to visualize the statistics.

Java File Structure

1. MainActivity

   • Registers/unregisters event listener
   • Records readings from the accelerometers
   • Calls RecognitionActivity for prediction and chooses activity with highest confidence score
   • Inserts predicted activity to a SQLite database in the background

2. RecognitionActivity

   • Initializes trained CNN classifier
   • Feeds normalized input into the classifier and sends output back to MainActivity

3. DisplayStatsActivity

   • Queries the SQLite database to display pie chart on the screen

4. Activity

   • Contains database schema in the form of objects (uses Room Persistence Library)

5. ActivityDao

   • Provides interface to modify and query the database (through Data Access Objects)

6. AppDatabase

   • Initializes an SQLite database for the application

7. Constants

   • Holds constant values used in all class files

Screenshots