Real-time Sinhala Writing Assistant for Kids

This paper presents a system that provides self-learning aids for children aged 5 and 7 years old to practice writing the Sinhala alphabet correctly. The system determines the correctness of the written letters based on the correctness of the shape of the letter, pen up-down movements, and writing the letter in the correct regions of the five lines provided, through the real-time letter writing correction module that utilizes the principles of online handwriting recognition approaches. Users are allowed to write the letters on a touch-enabled device and the real-time evaluation of the written letters is conducted using the template-matching. The proposed system assists the children to correct the errors in handwriting through a graphical representation. Furthermore, using a Convolution Neural Network (CNN), offline handwriting recognition is used to determine whether the child has written the same letter displayed by the system. The real-time letter writing correction module has an average accuracy of 94.28%, while the offline handwriting recognition module has an average accuracy of 95.0%.