Submitted by Automa??o e Estat?stica (sst@bczm.ufrn.br) on 2017-02-13T19:22:38Z No. of bitstreams: 1 CarolineAlbuquerqueDantasSilva_DISSERT.pdf: 1138216 bytes, checksum: b1401c36a2ad5415e6adc770fee68fbc (MD5) / Approved for entry into archive by Arlan Eloi Leite Silva (eloihistoriador@yahoo.com.br) on 2017-02-14T17:45:51Z (GMT) No. of bitstreams: 1 CarolineAlbuquerqueDantasSilva_DISSERT.pdf: 1138216 bytes, checksum: b1401c36a2ad5415e6adc770fee68fbc (MD5) / Made available in DSpace on 2017-02-14T17:45:51Z (GMT). No. of bitstreams: 1 CarolineAlbuquerqueDantasSilva_DISSERT.pdf: 1138216 bytes, checksum: b1401c36a2ad5415e6adc770fee68fbc (MD5) Previous issue date: 2016-07-18 / Coordena??o de Aperfei?oamento de Pessoal de N?vel Superior (CAPES) / Esse trabalho prop?e um esquema de otimiza??o convexa, baseada em programa??o linear e algoritmos gen?ticos, para equalizadores cegos aplicados a sistemas de comunica??es digitais. Ele surgiu da necessidade crescente de melhorias nos sistemas de comunica??o no intuito de transportar o m?ximo de informa??o poss?vel por um meio f?sico de forma con??vel.O esquema proposto, ELC-GA (Equalizador Linear Cego baseado em Algoritmos Gen?ticos), ? caracterizado por realizar a equaliza??o adaptativa cega do canal em blocos ?xos de dados, utilizando como algoritmo adaptativo um algoritmo gen?tico, cuja fun??o objetivo ? uma fun??o linear com restri??es, globalmente convergente. Entretanto, devido ?s caracter?sticas aleat?rias do sinal modelado com interfer?ncia intersimb?lica e ru?do aditivo branco gaussiano, a fun??o linear utilizada passa a representar uma programa??o linear estoc?stica. Nesse sentido, o uso de algoritmos gen?ticos ? particularmente adequado por ser capaz de buscar solu??es ?timas percorrendo uma por??o consider?vel do espa?o de busca, que corresponde aos v?rios cen?rios estoc?sticos. O trabalho tamb?m descreve os detalhes de implementa??o do esquema proposto e as simula??es computacionais realizadas. Na an?lise de desempenho, os resultados do ELC-GA s?o comparados aos resultados de uma das mais tradicionais t?cnicas de equaliza??o cega, o CMA, utilizado como refer?ncia dessa an?lise. Os resultados obtidos s?o exibidos e comentados segundo as m?tricas de an?lise adequadas.As conclus?es do trabalho apontam o ELC-GA como uma alternativa promissora para equaliza??o cega devido ao seu desempenho de equaliza??o, que atinge a converg?ncia global num intervalo de s?mbolos consideravelmente menor que a t?cnica usada como refer?ncia. / This paper proposes a convex optimization scheme based on linear programming and genetic algorithms for the blind equalizers applied to digital communications systems. It arose from the growing need for improvements in communication systems in order to transmit as much information as possible in a physical environment reliably. The proposed scheme, ELC-GA (Blind Linear Equalizer Linear based on Genetic Algorithms), is characterized by performing blind adaptive channel equalization in fixed units of data, using a genetic algorithm as adaptive algorithm, whose objective function is a globally convergent constrained linear function. However, due to the random characteristics of the signal modeled with intersymbol interference and additive white Gaussian noise, the used linear function now represents a stochastic linear programming. Accordingly, the use of genetic algorithms is particularly suitable for being able to get optimal solutions covering a considerable portion of the search space, which corresponds to the various stochastic scenarios. This work also describes the implementation details of the proposed scheme and the performed computational simulations. In the performance analysis, the ELC- GA results are compared to the results of one of the traditional blind equalization techniques, CMA, used as reference in this analysis. The results are shown and discussed under the appropriate metric analysis. The conclusions of the study indicate the GA - ELC as a promising alternative to blind equalization due to its equalization performance, which reaches global convergence in a considerably smaller range of symbols than the technique used as reference.