摘要
基于视觉的手势识别是实现新型人机交互的一项关键技术,针对手势普适性与识别率问题,在改进隐马尔可夫模型学习机制的基础上提出一种新的基于Leap Motion传感器的自适应动态手势识别方法。该方法首先采用几何特征法识别静态手姿以确定动态手势起始点与结束点,然后基于角度对动态手势轨迹进行特征提取与分类,并引入修正的重估方法计算隐马尔可夫模型参数,最后在对非定义手势识别的基础上自动学习更新隐马尔可夫模型,以提高动态手势识别率,并最终实现对26个小写字母的动态手势识别。实验结果表明,所提出的动态手势识别方法具有良好的自适应性与精确性。
Gesture recognition based on vision is a key technology to realize new human-computer interaction. To the problem of gestures adaptability and recognition rate,a new method of adaptive dynamic gesture recognition based on the Leap Motion is presented on the basis of improving Hidden Markov Model（ HMM）. Firstly,by using the method of geometric features to recognize the static hand posture,the start and end point of dynamic gesture trajectory could be confirmed. After extracting and classifying gesture trace features by angle,the revised revaluation method is introduced to calculate model parameters. Finally,with recognizing the undefined gesture,the method of automatic learning and updating HMM is presented to improve the efficiency of the gesture recognition,and the dynamic gesture recognition of 26 lowercase letters is realized in the end. Experimental results show this method has a good adaptability and accuracy performance in dynamic gesture recognition.
出处
《计算机应用与软件》
2017年第2期198-202,213,共6页
Computer Applications and Software
基金
上海市国际科技合作基金项目(12510708400)
上海市自然科学基金项目(14ZR1419700)
2015上海大学电影学院高峰学科项目(N13A30315W23)
