Spiking Neural Networks for Ultra-Low Power Radar Applications

Leuven - Master projects/internships
More than two weeks ago

Explore the use of Spiking Neural Networks for future cognitive edge applications deploying IMEC sensor technologies

Over the last decade, an increasing number of sensor systems have been paired with machine learning (ML) algorithms in order to enable and support an ever growing application space such as self-driving cars or object recognition hardware. Currently, most of these systems rely on cloud-based ML solutions especially for the learning aspect of the smart system.

While this approach can support a broad range of applications, it reaches its limit when the target application relies on decisions with low latencies such as in real-time systems or on decisions based on an extensive set of data while providing only a limited communication budget in terms of data rates. In additions, most of the times the same set of applications requires an execution with a limited energy budget as they are deployed in battery powered devices. A potential solution to this problem are spiking neural networks (SNNs) due to their event-based computation scheme.

Imec is working on the development of ML algorithms for the cognitive edge. In this context, we leverage on IMEC's knowhow in sensor technology (e.g. CMOS radar systems) and novel machine learning approaches such as SNNs.

In this topic, we propose to evaluate an SNN-based implementation for an application within the cognitive edge domain e.g. gesture recognition using a 140GHz CMOS radar system. This topic includes the following tasks:

  • Understanding of the sensor output data and its restriction on the machine learning algorithm
  • Implementation of a spiking neural network for the targeted application
  • Optimization of the spiking neural network in terms of size and targeted accuracy
  • Benchmarking of the implementation against a reference system (e.g. a traditional convolutional neural network) in terms of accuracy, size, latency, number of operations and other performance metrics

Type of project: Combination of internship and thesis

Duration: 6 to 9 months

Required degree: Master of Engineering Technology, Master of Engineering Science

Required background: Electrotechnics/Electrical Engineering

Supervising scientist(s): For further information or for application, please contact: Matthias Hartmann (Matthias.Hartmann@imec.be)

Allowance only for students from a non-Belgian university

Share this on


This website uses cookies for analytics purposes only without any commercial intent. Find out more here. Our privacy statement can be found here. Some content (videos, iframes, forms,...) on this website will only appear when you have accepted the cookies.

Accept cookies