A Virtual Keyboard Identification Algorithm Based on Keystroke Sequence Forecast
In the past research work, the keystroke information in the images can be identified. However, due to splitting the continuous process of keystroke into separate images, the original identification algorithm cannot distinguish the key pressed from the keys swept during finger movement exactly. The key blocks swept by the fingers are identified as the real keystrokes as well. Not all keystroke information identified is real after the single image recognition. Actually, it includes a lot of useless keys swept unintentionally during the keystroke process. In this paper, we propose a virtual keyboard identification algorithm based on keystroke sequence forecast It is regarded the keystroke process, including entering, pressing, and exiting, as a whole one. Furthermore, the user must exit the keyboard area before another keystroke process. A keystroke process begins with a blank image and ends with the next one. It analyzes images one after another with the former identification algorithm to get a keystroke sequence. It identifies the real keystroke by checking the change of keys weight in the sequence or comparing the number of different keys occurrence. In order to avoid the leakage of shoot, we use the keystroke video stream as input and get the images on a certain frequency. At last, an experimental verification with a 4x4 virtual keyboard is proposed. Compared with the former experimental results, the identification algorithm, a virtual keyboard identification algorithm based on keystroke sequence forecast, presented in this paper improves performance effectively.
virtual input method digital image processing keystroke identification algorithm keystroke sequence forecast
Jin Zhang Xiaowei Liu Heyang Pang Yang Li
Dept.of Computer Science Nankai University Tianjin,China
国际会议
太原
英文
381-385
2011-02-26（万方平台首次上网日期，不代表论文的发表时间）