What you will do

Millimeter-wave radar is increasingly used for contactless sensing and monitoring because it can operate independently of ambient illumination and directly capture range, velocity, and motion information. However, conventional commercially available MIMO radars have limited angular resolution, restricting their ability to reconstruct detailed spatial representations of pedestrians.

Inverse Synthetic Aperture Radar (ISAR) provides an alternative approach to high-resolution imaging by exploiting the relative motion between a target and a stationary radar. Previous work at imec has demonstrated a MIMO-FMCW ISAR pipeline for reconstructing high-resolution radar images of pedestrians. The approach combines target tracking, motion compensation and autofocus, time-window selection, MIMO phase compensation, and image formation. Experimental measurements with a 60 GHz radar demonstrated pedestrian imaging at distances from approximately 6 to 11 metres.

Despite these advances, high-quality ISAR reconstruction remains challenging because human motion is complex and non-rigid. In particular, imperfect motion compensation can introduce phase errors, artefacts, and image defocusing. Recent developments in deep learning and deep unfolding offer promising ways to combine radar-domain knowledge with trainable components for improved image formation and autofocus.

In this internship, you will investigate machine-learning techniques for high-resolution radar imaging of pedestrians, building on imec's existing FMCW/MIMO ISAR processing framework.

Your research activities

• Familiarize yourself with FMCW/MIMO radar signal processing, range-Doppler processing, tracking, and ISAR image formation.
• Process and analyse measured radar data from pedestrians performing different motions.
• Investigate the limitations of conventional autofocus and motion-compensation techniques for human targets.
• Develop machine-learning methods that improve one or more stages of the radar imaging pipeline, with emphasis on autofocus, motion compensation, image reconstruction, or image enhancement.
• Explore physics-informed, model-driven, or deep-unfolding approaches that combine known radar processing models with trainable components.
• Investigate training strategies that combine simulated radar returns with experimental measurements.
• Quantitatively compare the developed techniques with conventional signal-processing baselines using suitable image-quality and reconstruction metrics.
• Validate the resulting algorithms on experimental pedestrian radar measurements.

Possible extension: Depending on progress, the work may be extended toward 3-D radar imaging using multichannel or interferometric ISAR concepts. The core objective remains machine-learning-based autofocus and image reconstruction for pedestrian ISAR.

What we expect from you

• You are pursuing an MSc or PhD in electrical engineering, telecommunications, computer science, artificial intelligence, signal processing, or a closely related field.
• You have a strong interest in radar signal processing, machine learning, or computational imaging.
• You have experience with Python and preferably a deep-learning framework such as PyTorch.
• Knowledge of digital signal processing, Fourier transforms, estimation theory, optimisation, or array processing is highly valuable.
• Previous experience with radar, synthetic aperture imaging, computer vision, or complex-valued neural networks is an advantage but is not mandatory.
• You are comfortable working independently on a research-oriented problem and evaluating algorithms experimentally.
• You have good written and spoken English.

Research outcome

The internship will offer hands-on experience at the intersection of radar sensing, computational imaging, and machine learning. The objective is to develop and validate new algorithms capable of extracting significantly richer spatial information about pedestrians than conventional radar processing. Subject to the achieved results, the work may provide a basis for a scientific publication or a follow-up research project.