Utilização do vetor perceptually important points como Entrada para classificação de padrões com rede neural Artificial multi layer perceptron em série temporal

Abstract

Pattern Recognition in Time Series (RPST) is one of the fastest growing computing sub-areas in the world. Its exponential development has established the crossover among other areas, such as neuroscience, through various applications and solutions in the medical field, teaching and research. Although RPST has advanced in the neurosciences, a system with TPRS that recognizes and eliminates electroencephalogram artifacts has not been developed yet. In this context, the objective of this study was to develop a system with the use of RPST to identify, analyze and eliminate electroencephalographic signal artifacts. For this proposition, it was used the identification of critical points, that is, valleys and peaks that make it possible to identify a behavior standardized by the RPST, using the rule of perceptually important points (PIP). Then, electroencephalographic signal data were randomly selected with artifacts of eye blinking. The results demonstrate that the PIP is a comparative metric of RPST that can be used as input in Artificial Neural Networks (RNA) to categorize the pattern of the electroencephalogram spectrum, presenting an accuracy rate of 98.8% regarding the accuracy of artifacts elapsed along the electroencephalogram. In conclusion, the angular vector PIP use as a solution for identification in time series is applicable and accurate.