Smart sensors are becoming increasingly important to support applications such as autonomous driving, indoor monitoring, gesture or motion recognition and site surveillance, to name a few. The general approach for these applications is to rely on identical or heterogeneous sensors for detection, tracking and classification and to fuse the information at different levels to improve the accuracy and reliability of the information. IMEC has started several years ago the development of mm-wave radars in the E-band for such applications and is constantly improving them and also develops radars at frequencies higher than 100GHz.
In order to bring these radars closer to the applications and to make them smarter, imec is looking for a creative, autonomous and highly skilled PhD candidate who will develop algorithms in the following areas:
- Feature extraction and classification of high range resolution and micro-Doppler radar signatures and from cameras;
- Fusion of information (position, detection, tracking, classification, terrain, clutter) from several sensors illuminating/observing the same scene;
- Fusion of information from heterogeneous sensors such as radar, camera and LIDAR.
Very importantly, these algorithms will have to be at the same time highly performant and suited for real-time, embedded implementations with a reasonable power consumption budget and physical form-factor. It will be a part of the PhD objectives to demonstrate the developed algorithms on realistic platforms and in realistic environments.
The candidate will use techniques from the following scientific areas: stochastic signal processing, information theory, estimation theory, pattern analysis and machine intelligence. Very specifically, the selected candidate will have to demonstrate a strong background in several of the following domains:
- Advanced signal processing (filtering, FFTs, time-frequency transforms, cepstral analysis, array processing, wavelets, ...);
- Stochastic signal processing and system identification, Kalman filters;
- Estimation techniques (Bayesian estimation, least-squares, spectral estimation, super-resolution algorithms);
- Machine learning, supervised and non-supervised classification, neural networks, deep learning;
- Sensor technologies (radar, camera);
- Matlab, C/C++, Python;
- Real-time, high performance computing platforms.
The candidate will work in a team and will hence have to show good abilities to collaborate and communicate with other researchers. Excellent oral and written English are mandatory.
Type of work: algorithm development, Matlab/C modelling, implementation on real-time platform, demonstration in the lab
Supervisor: Hichem Sahli
Daily advisor: André Bourdoux
The reference code for this PhD position is SE1712-09. Mention this reference code on your application form.