Vision-based typing interface for chording input with keystroke detection using kalman filter
Poorna Talkad Sukumar, Kanchi Gopinath · 2013
In this paper, we present a vision-based typing interface implemented for chording input. Chording input is an interaction technique which uses combinations of finger touches to enter characters or commands. With this technique, we can obtain 127 combinations by using 7 fingers which is greater than the number of keys on a standard keyboard. But chording input requires identification of the fingers by some means. We describe the prototype system built which performs real-time recognition of fingertips and detection of keystrokes for a bi-manual typing interface using a single web cam. By imposing a few feasible constraints, we were able to use basic image processing for the detection and labeling of the fingertips. We also implemented a novel method for keystroke detection by velocity sensing using Kalman filter. Multi-touch detection is also feasible in the system and a Hindi/Devanagari text entry application has been implemented by encoding Unicode representations of the Devanagari character set into combinations of typed fingertips. The system achieved reliable fingertip recognition and keystroke detection with a mean latency of 80 ms.