Current systems in Weibel transmit a continuous signal with a constant frequency.
Any object in the beam will reflect the incoming wave with a frequency
shift that is proportional to its velocity and the carrier frequency (Doppler
shift). Typically, the radar observes scenes where only a few objects are
moving. Hence, the resulting measurement consists of only a few frequencies
and the resulting signal is sparse in the Fourier domain.
By the Nyquist criterion for Fourier analysis, it is
necessary to sample this signal at a high enough rate to reveal all its
frequency content. In contrast however, for such sparse signals, the theory
of Compressed Sensing promises the possibility to reconstruct the signal in
a near perfect way through sub-Nyquist sampling. This possibility allows to
lift pressure on the hardware coming from the high sampling rate.
Compressed Sensing is a powerful mathematical tool for
any form of signal (speech, radar, images) which has a sparse
representation in a basis.
The goals of this project are as follows. The student(s) will/should:
- learn and understand the mathematics of compressed sensing.
- do a literature study
- learn and understand when compressed sensing can be applied and to what purpose(s) it can be used. Especially, why and how it is possible o apply this theory in certain radar situations.
- implement their own compressed sensing algorithm in e.g. MATLAB, possibly using existing functions.
- Apply the algorithm to simulated and real radar data in order to test, verify, and find limitations of the compressed sensing algorithm.
Mads Sielemann Jakobsen
Phone: +45 4816 9343
Place of work: Weibel HQ and/or place of study.
Birte Kronbak Andersen is not the
project supervisor, she just put the project idea in the Project Bank
for Weibel. But a supervisor can most likely be found at DTU Space.
The student should have knowledge of minimization algorithms and of the Fourier Transform.