Ford Improved Shorthand Alphabet Recognition Based on Multi-Layer Perceptron in Neural Networks Using Java

Abstract

Every business organization is rapidly undergoing a digital transformation. In this transformation, paper documents from the past are being converted into a kind of format in which it is more efficient, saves more time in management and more secured. Optical Character Recognition(OCR) software can help extract valuable information from the scanned documents and store the data in a way that it is readily available. OCR converts scanned paper documents and digital files into editable and searchable data. The researchers used a multi-layer perceptron in a neural network to implement optical character recognition. Furthermore, the researchers created a software that will recognize handwritten letters of Ford Improved Shorthand Alphabet from images. The results of the study showed that many factors affect the accuracy of software/program in recognizing letters from the inputted images, including the lightness or darkness of the strokes for recognition, the thickness or thinness of the handwritten letters in the images, and amount of trained neural network for recognition. Therefore, the larger the amount of data trained for each letter of the Ford Improved Shorthand Alphabet, the more accurate the recognition of the software/program will be. Furthermore, the clearer the inputted images for training neural networks as well as the coded models for the identification, the accuracy of the software/program varies greatly.