Alternativas de redução de ruído e exploração de algoritmos de filtragem adaptativa em receptores para canais seletivos em frequência.

Abstract

This work deals with the design of techniques to combat the effects of additive noise and intersymbol interference (ISI) on receivers, for frequency-selective fading channels. Those techniques are suitable for blind reception and training sequence aided reception. Concerning the blind reception, this thesis focuses on receivers that use the maximum likelihood criterion to detect symbol sequences employing per-survivor processing (MLSE-PSP), for mobile communications channels subjected to time-variant frequency selective fading (TVFSF). Special attention is given to the choice of the adaptive filtering algorithm used by the receivers. The analysis of the performance of those receivers shows that they are severely affected by burst errors in the detection processing. In order to overcome this problem, a procedure is implemented to detect the occurrence of errors in digital receivers. This procedure is based on a new concept introduced in this thesis: filtering diversity. This procedure is computationally simple, works in time-variant channels, presents good performance and can be used for any reception scheme. Other important theme treated in this thesis is the analytical evaluation of the steady-state behavior of the adaptive filtering algorithm LMS {Least Mean Square) on the identification of channels subjected to the effect of TVFSF. Closed-form expressions are obtained for the mean squared error (MSE) and for the optimum step of the LMS. Relating to the processing to combat the effect of the additive noise, the use of denoising techniques is proposed as an alternative method to the conventional schemes, which use linear time-invariant filters. The denoising techniques are implemented in the Wavelet transform domain and they take advantage of the parsimonious property of this transform. The performance of the proposed front-end scheme is analyzed considering two aspects: the analytical one, for AWGN channels, and the use of simulation, for frequencyselective channels. In the first case, approximations are obtained for the system bit error probability in terms of the parameters of the Wavelet transform and the denoising technique. For frequency-selective channels, many configurations of decision-feedback equalizers are analyzed. It is observed that the proposed scheme outperforms the conventional ones, mainly for the case when the channel frequency response presents notches in the frequency band of the transmitted signal.