基于小样本学习的3D动态视觉手势个性化交互方法

Personalized Interaction Techniques of Vision-Based 3D Dynamic Gestures Based on Small Sample Learning

传统的动态手势交互技术如隐马尔科夫模型、神经网络和统计分类器等都需要大量的训练样本,建模过程中需要领域专家的干预、对普通用户来说使用起来较为困难,并且它们针对的是特定的手势集合,很难对其进行扩展.本文通过WOZ实验,分析了用户的行为特征并给出了基于手势的数字电视交互任务模型;提出了3D动态手势状态转移模型,解决了Midas Touch问题;提出了一种基于小样本学习的动态手势识别方法,解决了传统手势识别方法的缺点;构建了个性化手势设计平台,满足了用户的个性化定制需求;实验评估结果验证了本文方法的有效性.

There are some unresolved issues left behind for many traditional dynamic gesture recognition methods ,such as Hidden Markov Model（HMM） ,Neural Network（NN） ,and statistical classifiers .For example ,they require a large number of train-ing examples and the involvement of expert users in the training process .Moreover ,they are used for some specific gesture sets which are difficult to be extended .In this paper ,we first build a task model and a state transition model for vision-based dynamic gestures .Then we propose a method for 3D dynamic gesture recognition based on small sample learning .Next we design a toolkit for development of user-defined gestures .Finally ,we develop a gesture-based interactive television prototype .Experimental results verify the validity of our method .