Digital Pen for Handwritten Alphabet Recognition

A novel technology for the digitalization of the human handwriting system is studied in this paper. A recognition algorithm based on image processing using a scanner or touch panel has been used as a typical way to digitize handwritten characters. However, a new digital pen proposed in this study can recognize the handwritten characters in real time. The force information applied at around the tip of a pen, and accelerometer and gyro sensor during writing is utilized for character recognition. A new pen device including multiple tiny force sensors was designed to measure the force vector of the pen tip. The force data of handwritten lowercase alphabets was prepared for convolutional neural network classification. The data is preprocessed by using a shift window, and rearranged as a dataset of a certain size, and used for training and evaluation of the character recognition algorithm. The recognition result exhibits a high accuracy character classification of all alphabet.