| dc.description.abstract | The main purpose of this research is to create a fully functioned system to translate any handwritten mathematic expression into LaTeX code. This research itself serves as one of the basic part of the system, the handwritten character recognition system. Three feature extraction methods were compared and evaluated. They are Feature Point Extraction, Principle Components Analysis, and Fast Fourier Transform. Classification method used in this research is K-Nearest Neighbors. Accuracy measurement of the three methods shows that the maximum accuracy score by Feature Point Extraction is around 26%, while Principle Component Analysis and Fast Fourier Transform score is approximately 60% and 70%, respectively. FPE, despite its high score on optical character recognition (around 86% accuracy score), did not perform well due to the fact that the FPE method used in this research did not aware of the position of each feature point. PCA and FFT proved to be better for handwritten character recognition, with FFT being the one to have the highest accuracy score. | en |
