A virtual character recognition system based on optical detection of red light and its embedded implementation

This paper proposes a novel ldquoair-writingrdquo character recognition system (ACRS) based on optical detection of red light, with which a user can write a character in the air with a red light-emitting device. The trajectories of the light can be captured and detected by a camera during the writing process and then a character reconstruction algorithm is employed to convert them to a 2-D plan (as a virtual character trajectory) after a serials of preprocessing steps, including background analyzing and noise erasing. Finally, the virtual character trajectory is considered as a kind of inkless character pattern and recognized by a minimum classifier. To expand the application of our system, we implement an ARM7-based embedded ldquoair-writingrdquo character recognition system (E-ACRS). Experiments show that both ACRS and E-ACRS are able to recognize cursive digits and English alphabetical characters with encouraging accuracy.