Using FPGAs for Spectrum Sensing and Modulation Recognition

Project Objective

This project seeks to use machine learning to recognize different wireless devices. The project will use software defined radios (SDR) to record various devices, such as iphones, bluetooth earbuds, and Wifi laptops. These recordings will become the training data to a set of neural networks. The second part of the project will quantify the accuracy of the neural networks in classifying various device types based on their RF signatures.

Reading Material

• ​Overview of FPGA architecture (especially for Xilinx devices), and comparison between FPGA and CPU
• ​Overview of I/Q Communication Theory
• ​Artificial WiFi Packet Generation
• ​Matched Filter Demodulation

Week 1 Activities

• Get ORBIT/COSMOS account and familiarize oneself with the testbed procedures
• Learn about FPGAs
• ​Presentation 1

Week 2 Activities

• Write a UDP client / server in Go
• Generate IQ samples for IEEE 802.11 WiFi packets
• ​Presentation 2

Week 3 Activities

• Rework UDP client / server to work with Go to Verilog compiler
• Transmit and receive generated WiFi packets using the USRPs on the Grid
• ​Presentation 3

Week 4 Activities

• Automate data collection on the Grid
• Learn more about Go lang
• ​Presentation 4

Week 5 Activities

• Begin looking in to matched filters
• Finish data collection on the Grid (a lot of debugging)
• ​Presentation 5

Week 6 Activities