Project information

---

• Type: Software | Deep Learning
• Skills: Python | Aanaconda | Keras | Git | Deep Learning | BLE | Firmware
• Role: Undergraduate Research Assistant
• Code: https://github.com/UrbanPistek/BLE_Sensing

The project objective was to develop a system for detecting human presence using BLE technology. The ruvvi tags were used as the BLE device and the ruvvi developer sheild was used to flash the ruvvi tags special firmware using the Nordic NRF5 SDK. Large sets of RSSI data was collected from the ruvvi tags which was then processed using a combination of excel and python. The data was then fed into a custom developed Convolutional Neural Network design to extract the most important features from the RSSI data and then classify whether a person, multiple people or no people are present in the room at that current timestep. Through testing and optimization the model was able to achieve a 80% prediction accuracy on out of sample data.