Handwritten Character Recognition Using Convolutional Neural Network and Contour Detection
ABSTRACT
Rapid technological advancement has made world going more digitally in most aspects. Despite that, the current technological advancement does not make people leave handwriting as their daily activity. Handwritten recognition is easily done by humans, but we need a machine or computer taking over the job when it comes it recognizing a lot amount of handwritten data. The goal of this paper is to find the best parameter for character recognition models. This paper investigates the best parameter to predict a handwritten character based on Convolutional Neural Networks (CNN). The parameter analysis will be completed by the analysis of variance (ANOVA) test. Other than that, we tried to predict a word using Contour Detection for segment a word into a single sequence character. This study shows that the best accuracy and the least training computation time is 98.42% and 298.84 seconds respectively. However, the systems only reach 72% accuracy when it comes to recognizing a word.
#ConvolutionalNeuralNetwork #ANOVA #ContourDetection