Machine Learning for Interference rejection in Body Area Networks

Introduction

NXP Semiconductors provides High Performance Mixed Signal and Standard Product chips/IC’s to deliver secure connections for a smarter world.

The NXP Product Line Personal Health, headquartered in Leuven, develops application specific IC’s for use in “Hearing Aids and Implants”, “Wireless Audio” and “Wearable Health monitoring” applications.

Future Wearable Health monitoring applications may make use of several body worn devices, communicating wirelessly. The goal is to investigate neural networks for interference rejection capabilities and spectral awareness in a wireless body area networked communication link.

Motivation

There are several wireless communication technologies that can be used to create a Wireless Body Area Network (WBAN) and they could operate at different frequency bands. Some examples of this are at very low frequencies around 10MHz and other examples operate at ISM bands like 864MHz, 2.4Ghz etc.

Irrespective of which frequency bank WBAN operates, unwanted interfering signals can significantly degrade the communication link performance of a WBAN. Finding methods to mitigate these interferers is essential to ensure a robust communication link

Work

Signal separation or interference mitigation will help in extracting the required signal of interest and thereby improving the fidelity and performance of the communication link. Source separation activity using Machine Learning (ML) techniques have demonstrated signification performance improvements in other domains link vision, audio, etc.

The goal here is to apply machine learning techniques to RF signals, where the wanted signal of interest and any unwanted signals are identified and separated. The work will involve setting up a ML environment to train a ML model with database(s) comprising of wanted communication signals along with relevant interference signals. The student should also study / look into wider ML challenges taking place in this domain, where training data bases are available. This can be a good starting point to setup the ML environment and start the investigation of algorithms.

This work is planned for around 5 months full-time internship at NXP.

Deliverable The result of the work will be to come up with a ML framework / environment and the required algorithms, ML models that will demonstrate interference rejection capability in a WBAN communication link.

Profile of the student :

You study as master in electrical engineering, nano-technology or information technology. You are interested in wireless communication, machine learning, and have notions of ML frameworks like Tensorflow, Pytorch, Matlab etc. You are creative, hands-on and have good communication and English language skills.

Work environment :

You have the opportunity to work in a fully equipped high tech environment. You will be coached by experienced engineers. Your location is mainly NXP Leuven.

Master

Computer Science Engineering, Electrical Engineering

Sector

Electronics

Locatie

Flemish Brabant

Stad

Leuven

Voordelen

learning on the job

Gezocht Profiel

Profile of the student :

You study as master in electrical engineering, nano-technology or information technology. You are interested in wireless communication, machine learning, and have notions of ML frameworks like Tensorflow, Pytorch, Matlab etc. You are creative, hands-on and have good communication and English language skills.

Mail