Résumé:
In this thesis we have discussed the estimation of the multipath fast fading
channel in OFDM systems. Channel estimation requires a large number of
pilots, which reduces the number of subcarriers dedicated to the useful data
transmission. In this thesis, our main goal is to carry out channel estimation
algorithms aiming to achieve the best combination of estimation performance
and maximum spectral efficiency, using the basis expansion model (BEM).
In particular, we focus on the development of a reduced PDR (Pilot to useful
Data Ratio) estimation techniques by proposing two techniques in order to
improve the spectral efficiency while maintaining an acceptable estimation
performance in terms of Mean Square Error (MSE), and Binary Error Rate
(BER), in high mobility situations. Our contribution is mainly based on
exploiting additional clustered pilots carried by adjacent OFDM symbols,
in complement of those contained in the symbol under investigation. The
proposed adjacent pilot-based method leads us to consider the two followings
cases, according to the assumed channel hypothesis :
1) In the first case, we consider a doubly selective channel, based on a
comb pilot structure with the BEM (basis expansion model) approximation. In such scenarios, certain statistical properties of the channel
are assumed to be a priori known.
2) In the second case, we consider a high mobility situation along with
another type of channel, known as multipath channel with sparse taps.
This type of channel exhibits a fast variation of the taps delay interspacing, which causes a very large delay spread between the different
taps.
