Abstract:
The objective of this project is to improve the performance of adaptive acquisition of
PN sequences in DS/CDMA systems. This latter suffers from rapid variation of noise level
and propagation conditions. Therefore, fixed threshold acquisition systems are unable to
adapt to this type of environment and can lead to a high rate of false alarm and/or low
probability of detection. Consequently, we propose an adaptive arrangement of the
detection threshold, through the use of a constant false alarm rate, well known in radar
detection. In this thesis, two adaptive acquisition systems are proposed.
The first system uses the double-dwell search strategy and the automatic censoring
technique to annihilate the shadowing effect caused by the presence of multiple paths in
the reference channel. This system uses a smart antenna at the receiver. The performances
of the proposed systems, in terms of detection probability and mean acquisition time, are
analyzed in the absence and in the presence of multipath and multiuser signals. Then, they
are compared with several systems proposed in the literature. The obtained results show
the robustness of this system, especially in the presence of multiple access interferences.
In addition, it reduces considerably the mean acquisition time.
The second system is based on the serial search strategy with its simple structure. It
uses a detector made by multilayer artificial neural networks and trained by the error retropropagation algorithm. Its performances are analyzed and evaluated in two types of
channels, Gaussian and Rayleigh, in the presence of multiple paths for various parameters.
The obtained results show that the neuronal detector improves significantly the probability
of detection and the mean acquisition time of the serial system; because it greatly reduces
the computation time of the acquisition process.
