Development of a Baybayin Words Recognition System Using Support Vector Machine

Ancient scripts like Baybayin hold immense cultural heritage yet remain inaccessible in modern digital applications. This is due to optical character recognition (OCR) systems struggling to recognize Baybayin characters accurately due to their rather complex structure. This research develops a Support Vector Machine (SVM) system that recognizes and converts Baybayin words into their respective Tagalog representation. The linear and RBF kernels enhanced character prediction, improving accuracy and robustness. An overall accuracy score of 96.88% was obtained for the RBF kernel on the more straightforward Baybayin characters and 93.59% for the Linear Kernel on the more complex Baybayin characters. The processed images are classified into central character bodies and accent signs; the characters are predicted and combined before changing into their respective character equivalents from the accents. Characters are recombined after accent modifications to form the predicted Baybayin word. Testing on a dataset of 120 sample words of varying words written in Baybayin achieved 90.83% accuracy in word recognition. This indicates SVM's capability to distinguish the intricate strokes and variations of the Baybayin script. The optical recognition system provides an advancement towards digitally archiving and revitalizing the Baybayin cultural heritage