一种基于嵌入式系统实时交互的手势识别方法

摘要

提出了一种在单摄像头条件下基于嵌入式系统的手势识别方法.通过拟合手势图的外接多边形,找出其所对应的手势缺陷图,并建立手势与手势缺陷图的一一映射,利用手势缺陷图的特征来匹配和识别不同的手势.算法还将手势的跟踪与识别有机地统一起来,通过预测下一帧中手势出现的粗略位置大大降低识别步骤的计算量.该算法在实际应用的嵌入式平台下,能快速、准确地实现手势的识别,能够满足实时人机交互的要求.%This paper proposed a novel real-time gesture recognition approach based on embedded system to capture the motion of hand gesture with single camera. The proposed approach found out its corresponding defect map by fitting the external polygon graphics gestures, and established gestures' and defect maps of mapping and then matched the characteristics of gestures using defect map. It also combined hand tracking with gesture recognition,reducing the amount of calculation by predicting the approximate position of hand in the next frame. Experimental results show that the proposed method can work efficiently and accurately under embedded platform, and thus it can meet the requirements from real-time vision-based human-computer interaction.