Estimation et égalisation des canaux sans fil épars utilisant l’algorithme des moindres carrés récursifs épars SPARLS

Abstract

SPARLS, a channel estimation sparse scheme, outperforms the conventional widely used RLS algorithm in terms of mean squared error (MSE) and computational complexity. This work aims to adapt SPARLS to perform sparse channel equalization and to evaluate its performances. Moreover, the implementation aspect is analyzed in order to propose a hardware solution which takes advantage of sparsity to meet the constraints of real-time processing for WCDMA receivers. In channel estimation, our evaluation criteria are mainly accuracy, convergence rate and computational complexity. Accuracy is measured through the evolution of the steady state error depending on the signal to noise ratio (SNR). However, convergence rate and tracking properties are linked to the temporal behaviour of error. The symbol error rate is considered as an additive criterion in the equalization context. It reveals robustness of the evaluated methods facing various channel distortion levels. As a channel estimator, SPARLS prove a significant improvement over RLS and maintains its superiority in terms of convergence rate compared to our newly proposed thresholding based variant of RLS. Regarding equalization, a theoretical study of predefined parameters of SPARLS is recommended to regain the same level of performance obtained in channel estimation. To manage the irregularity of sparse variables calculation, we proposed a partitioning of the implementation between a digital signal processor (DSP) and an FPGA; this solution provides the flexibility required to exploit sparsity while ensuring a reasonable cost-effectiveness.

Date	7 Dec 2012
Original language	French
Awarding Institution	École de technologie supérieure
Supervisor	Ghyslain Gagnon (Supervisor) & Claude Thibeault (Co-supervisor)