Detectores adaptativos robustos projetados via otimiza??o convexa para sensoriamento espectral em r?dios cognitivos sob ru?do impulsivo

• Autor(a) principal: Ribeiro, Carlos Francisco de Almeida Cavalcanti
• Data de Publicação: 2013
• Tipo de documento: Dissertação
• Idioma: por
• Título da fonte: Biblioteca Digital de Teses e Dissertações da INATEL
• Texto Completo: http://tede.inatel.br:8080/tede/handle/tede/38
• topic: CVX; detector adaptativo; detector minimax, detector minimax robusto; otimiza??o convexa; r?dio cognitivo; ru?do impulsivo; sensoriamento espectral por autovalores; Engenharia - Telecomunica??es

This master thesis proposes a new approach for the design of optimal adaptive detectors for centralized cooperative spectrum sensing based on the eigenvalues of the received signal covariance matrix. The project is treated as a convex optimization problem from which detectors are developed under the minimax criterion, adapted to the presence and absence of impulsive noise. Non-robust and robust detectors are analyzed. In the first case we know a priori the distribution of the test statistic, whereas the second assumes only the knowledge of some moments of this statistic. Four detection techniques are discussed: the energy detection, the detection by the ratio between the maximum eigenvalue and noise variance, the detection by the ratio between the maximum and minimum eigenvalues and the test of generalized likelihood ratio test. The design and performance evaluation of the detectors are made under the influence of impulsive noise. It appears that the performances of non-adaptive and adaptive detectors are similar in the absence of impulsive noise for all the techniques evaluated. However, it is observed that the adaptive detector offers superior performance under impulsive noise, with emphasis on the detection technique based on the generalized likelihood ratio test. The robust adaptive detectors, regardless of the detection technique, shows superior performance compared to non-robust ones under impulsive noise, with the additional advantage of not requiring a priori information about the distribution of the test statistic.