提出了一套基于指势的识别系统, 引入OPENCV计算机视觉库, 为系统软件的设计及实现提供了便捷的代码。不同于以往方法依赖于正对摄像头(垂直)投影的限制, 该系统能够在任意倾斜角度情况下进行指势识别。其基本方法包含肤色检测、大量连续微移动(Consecutive Count of Micro Movements, CCMM)算法和图像降噪及轮廓提取的预处理部分, 然后运用k值法进行指尖及指蹼点的检测。最后, 由决策树算法识别出手指指尖和指蹼点构成的简单指势的特定含义。该系统运行稳定, 运行速率较快, 平均识别率高达95.05%, 并包含了参数调试功能, 适用范围较广。

A set of identification system based on the finger gesture recognition was put forward.OPENCV library provided a convenient code for system software. Unlike the previous methods with limitation on facing the camera, the system is able to identify tilted finger cases of any angles.Its basic method contains skin color detection, a large number of consecutive count of micro movements (CCMM) algorithm, noise reduction and contour extraction. Then, the k value method was used to identify the fingertips. Finally, the specific meanings of finger gestures were recognized by the decision tree algorithm. The proposed system runs stably and faster, and contains the function of parameter debugging with wider adaptation range as well. The average recognition rate could reach as high as 95.05%.