Deep Learning for Wireless Communications

Overview

Next generation wireless systems need to operate in harsh environments, where various types of interference increase the system level challenges. Wireless receivers have numerous applications in systems that require efficient spectrum management. In this session, we will demonstrate how to apply techniques Deep Learning and Machine Learning networks for a range of wireless communications systems.

We will look at the trade-offs between machine learning and deep learning workflows.  We will also demonstrate ways to perform data collection and labeling from off-the-shelf software-defined radios and radars to train and test classifiers. Our focus will be on data synthesis to train networks including efficient ways to work with communications baseband I/Q signals to improve classification results.

Highlights

Demonstrate the concepts and workflows using several application examples including waveform modulation ID, RF Fingerprinting, and 5G channel estimation

Understand trade-offs between machine learning and deep learning techniques for baseband signals

Pre-process and label baseband data

Synthesize data to train networks

About the Presenters

Dr. Houman Zarrinkoub is a senior product manager at MathWorks responsible for wireless communications products. During his 20-year tenure at MathWorks, he has also served as a development manager and has been responsible for multiple signal processing and communications software tools. Prior to MathWorks, he was a research scientist working on mobile and voice coding technologies in the Wireless Group at Nortel Networks. He has been awarded multiple patents on topics related to computer simulations of signal processing applications. Houman is the author of the book Understanding LTE with MATLAB: From Mathematical Modeling to Simulation and Prototyping. He holds a B.Sc. degree in electrical engineering from McGill University and M.Sc. and Ph.D. degrees in telecommunications from the University of Quebec, in Canada.

Florent Busnoult is a Senior Application Engineer at MathWorks focusing on signal processing and wireless communications products. He holds a Master’s degree in telecommunications engineering from Telecom Bretagne, in France.