中文摘要：

为满足便携式设备中人机交互的需要,设计了嵌入式手指交互系统。研究了系统所采用的肤色分割、凸包计算、指尖检测等算法,并完成了硬件设计。首先,根据肤色的聚类特征,在对比分析常用彩色空间特性的基础上建立肤色模型,对人手进行分割;提出射线扫描法对经典Graham扫描算法进行改进,快速计算人手凸点。然后,分析了利用手指轮廓弯曲特征检测指尖的算法。最后,介绍了以DSP和FPGA为微处理器构成的硬件系统。实验结果表明,设计的系统对自然伸展的单个手指正确检测率为95.2%,对弯曲手指的正确检测率为92.6%,对在非目标手指干扰下的正确检测率为90.1%;而对指尖的定位最大偏移量为2.12 mm;指尖定位总耗时约为23 ms。所设计的嵌入式手指交互系统稳定可靠、满足实时要求。

英文摘要：

To meet the need of human-machine interaction of the portable equipment,an embedded finger-interaction system was designed,and the algorithms applied to the system,such as skin color segmentation,convex hull computation,and fingertip detection,were investigated and the hardware was also implemented.Firstly,according to the clustering characters of skin color,a human hand was segmented via building the skin color model by analyzing the characters of common color spaces.Then,the classic Graham-scan method was improved based on the radial-scan method,and the convex hull of the human hand was computed quickly.Furthermore,the algorithm of fingertip detection which uses the curve of human finger was discussed.Finally,the hardware structure based on the Digital Signal Processing（DSP） and Field Programming Gate Array（FPGA） was introduced.Experimental results indicate that the detecting precision is 95.2% for a natural stretched finger,92.6% for the bended finger and 90.1% for other fingers under the disturbance,respectively.Moreover,the maximum offset of fingertip location is 2.12 mm and the location time is 23 ms.The system is natural and friendly,and has strong stabilization and real-time characteristcs.