Estimacao do numero de fontes via norma L0,55

Detalhes bibliográficos
Ano de defesa: 2016
Autor(a) principal: Floriano, Giovanni Henrique Faria lattes
Orientador(a): Guimar??es, Dayan Adionel lattes, Souza, Rausley Adriano Amaral de lattes
Banca de defesa: Guimar??es, Dayan Adionel lattes, Souza, Rausley Adriano Amaral de lattes, Pimenta, Tales Cleber lattes, Chaves, Felipe Emanoel lattes
Tipo de documento: Dissertação
Tipo de acesso: Acesso aberto
Idioma: por
Instituição de defesa: Instituto Nacional de Telecomunica????es
Programa de Pós-Graduação: Mestrado em Engenharia de Telecomunica????es
Departamento: Instituto Nacional de Telecomunica????es
País: Brasil
Palavras-chave em Português:
Palavras-chave em Inglês:
Área do conhecimento CNPq:
Link de acesso: http://tede.inatel.br:8080/tede/handle/tede/4
Resumo: In 2014, an empirical method was proposed to estimate the number of signal sources being impinged by multiple sensors, named norm-based algorithm (NB). This method is based on analysis of the Euclidean norm of vectors whose elements are the eigenvalues, standardized and scaled nonlinearly, from the received signal covariance matrix and the also standardized corresponding indexes. These norms are then used to discriminate the largest eigenvalues among all of them, allowing the estimation of the number of sources. In this dissertation we propose the improved norm-based algorithm (iNB) which uses the ???0,55-norm to discriminate the largest eigenvalues. Unlike the NB algorithm, iNB does not use the nonlinear scaling, and also does not require the definition of an empirical constant to perform the discrimination of the eigenvalues. To evaluate the effectiveness of this proposal (iNB) their results were compared with the results obtained using estimators based on Akaike information criterion (AIC), the method length minimum description (MLD) and random matrix theory (RMT). These comparisons show that iNB algorithm can overcome the performance of one or more of these estimators in various situations and that iNB always exceeds the NB Algorithm.