Conversão simbólica de sinais digitais por meio da Teoria de Extremos Relativos

AUTOR(ES)

Daniel Joaquim de Sousa

DATA DE PUBLICAÇÃO

2009

RESUMO

The goal is to develop a new technique for symbolic conversion of digital signals, called Quantization based on Relative Extrema (QBER). This technique can convert unidimensional digital signals into strings. The technique QBER, formalized in this proposal, uses Relative Extrema Theory (TER ) and signals similarity functions, as metric Edit Distance with Real Penalty (ERP). In addition, also uses the clustering algorithm PAMSLIM, which employs k-medoid approach, widely discussed in the literature. The TER is also other contribution of this work, as an extension of Important Extrema Theory, increasing concepts of prevalence, mount, Relative Extrema based Representation (RBER) and Quantized Relative Extrema based Representation (RBERQ). To evaluate the usefulness of QBER, has developed a classification system of reference patterns, based on the KNN classifier. This reference implementation has the pre-processing and recognition phases. In pre-processing, QBER is used to convert training objects of KNN in symbolic representations. As the k-Nearest Neighbor (kNN) classifier uses instance based learning, a training phase is inexistent, all classification is based on the training objects. In pre-processing phase also an object in analysis is converted to RBERQ symbolic representation, before serving as input for the classifier KNN. In order to evaluate the usefulness of the technique developed, are made comparisons in a problem of classification - generation of recommendations for the purchase of shares. The kNN classifier implemented is evaluated with and without the use of QBER, being useful in the employment by superior performance (of QBER) considering aspects of preparation time and annual return obtained.

ASSUNTO(S)

ciencia da computacao symbolic conversion of digital signals erp knn pam-slim conversão simbólica de sinais digitais

Documentos Relacionados

• A formulação dos 'estados relativos' da teoria quântica
• Solução exata de problemas de escalonamento determinísticos por meio de verificação simbólica de modelos
• Restauração de gravações fonograficas por meio de processamento digital de sinais
• O despertar para a leitura por meio de mídias digitais
• Avaliação da qualidade de diferentes técnicas de obturação endodôntica, por meio de imagens radiográficas digitais.