Bima script handwriting pattern recognition using histogram of oriented gradients and backpropagation classification method

The Bima script is one of the cultural heritage of the archipelago that needs to be preserved. Based on the results of a questionnaire conducted by the author online to 81 respondents from Bima, 48.1% had never studied the Bima script; moreover, 45.7% of people do not even know the existence of the Bima script. The dataset is collected from 20 respondents, which each responcent write 22 letters of the Bima script 12 times. The purpose of this research is to build a machine learning model that can recognize the handwriting pattern of the Bima script using the Histogram of Oriented Gradient (HOG) feature extraction and the Backpropagation classification method. This research results get 97.70% accuracy, 97.72% precision, and 97.65% recall, which used 1 hidden layer, 128 neurons, 0.5 dropouts, 1500 epochs, and a learning rate of 0.001 with an image size of 64x64 pixels.