LAUSR

Abstract Character recognition is a technology that digitizes data stored on paper for permanent archiving. Character recognition processes are divided into two steps: those that convert a character on paper as a digital image (scanning), and those that infer a scanned image as a character (recognition). Optical scanners and cameras are conventionally used for the scanning process. However, noise and distortion, which lower the recognition rate, can occur with such optical method. Herein, a triboelectric nanogenerator (TENG) is utilized for the scanning process. The TENG converts mechanical energy into electrical energy based on contact electrification. A single-electrode TENG with multiple aluminum cells was fabricated as a character-recognition TENG (CR-TENG). Using the difference in the intensity of the contact electrification, the aluminum cells of the CR-TENG distinguished ink-printed paper from bare paper. The CR-TENG has advantages in that it consists only of aluminum on the substrate and does not require additional electrical energy to operate. The CR-TENG converted numerical patterns on paper into images and a pre-trained deep neural network inferred the images as digits. Through optimization of the scanning and recognition process, a high recognition rate was achieved with this new character recognition system based on the CR-TENG.