摘要
为了解决手势控制服务机器人时单手手势交互不够自然和易受背景光照影响的问题,提出一种较自然的双手手势识别方案。利用Viola-Jones检测器和高斯肤色模型结合的方法进行手势分割,有效克服了背景和光照的影响。针对单一特征识别时的缺点,提取手势形状的多种特征组成混合特征向量。最后通过经典的BP神经网络方法实现手势识别。实验表明,该系统在不同背景和光照条件下鲁棒性较好,识别率在90%以上,可以对服务机器人进行前进、后退、左转、右转、停止的实时控制。
In order to solve the problems that using single hand interaction is not natural and be influenced by back-ground or illumination when the service robot movement is controlled by hand. A kind of more natural two-hand recognition scheme is proposed in the paper. Combine the Viola-Jones detector with Gaussian skin model for hand detection, the influence of background and illumination is overcame effectively. To make up for the disadvantage of single feature recognition, a variety of characteristics about the hand shape is extracted. Finally, through the classical BP neural network method the hand recognition is realized. The experimental results show that this system has a good robustness in different background or illumination, recognition rate achieves more than 90%. And the system can realize the real-time controlling of forward, backward, turning left, turning right, stop to robot.
出处
《电子测量与仪器学报》
CSCD
2013年第4期305-311,共7页
Journal of Electronic Measurement and Instrumentation
基金
安徽省科技攻关项目(1206C0805039)