UMA HEURÍSTICA GERAL PARA COMPARAÇÃO DE SINAIS
Pedro Jorge de Albuquerque de Oliveira1; Nadia Nedjah1; LUIZA DE M. MOURELLE1
1 Universidade do Estado do Rio de Janeiro
doi:10.20906/CPS/CILAMCE2017-1308
Resumo
The need to devise a general, methodical and objective comparison measure is both fundamental and ubiquitous, given the increasingly common application of techniques from statistics and machine learning to the solution of science and engineering problems. This kind of similarity measure has potential applications in fields as varied as system model identification, evolutionary systems design, computer vision, image processing and video processing. The reliable evaluation of members of a dataset, regarding their pertinence and distance to said set, can also be seen in the work of Mahalanobis. In this work, we propose a heuristic function of a general kind aimed primarily at enabling the comparison of two datasets and the quantification of the "distance" between them based solely on information that may be extracted directly from it using common methods from statistics and signal analysis. The idea is, fundamentally, that through the application of the similarity that we propose it should be possible to identify if a sampled signal is contained within a given datum. It was originally devised to enable the objective comparison and ordering of circuits designed based on their output responses, as part of a genetic programming based design procedure. In this paper, it is our aim to expose the merits and analyse the means of application, as well as the potential drawbacks of the proposed technique, we also intend to compare our heuristic function to other such metrics. It is our belief that the main contribution of this paper lies in the exposition of a comparison metric that simultaneously exploits the characteristics of a given signal and its Fourier spectrum, thus making it a robust technique for analysing sampled signals corrupted by random noise.
Palavras-chave: Análise de correlação; Heurística para a comparação de sinais; Avaliação de qualidade de modelos