Human Activity Recognition

An Android App named PDIoT which implements real time human activity recognition, which can classify up to 18 activities and can achieve 95% accuracy on 5 activities classification. The APP connects wearable RESpeck sensor via bluetooth. The project is a part of Principles and Design of IoT Systems Course I took in 2022, the dataset that we used to train this model was directly collected from 49 students who have taken this course. 

Below is how the app is designed and with functions of blue tooth connection, showing live data from the sensor and a sedentary reminder. You can find it in Github repo.

This demo shows classification results when a subject (me) falling, it has a 2 sec delay since we feed every 2 second data to the deployed TFlite model to get result.