一种中国手语单手词汇识别系统

利用数据手套，视觉设备和肘部弯曲传感器作为输入设备，提出了一种基于特征匹配和信息融合的中国手语单手词汇识别系统，首先利用手姿信息和手势信息的类型判别来确定待识词所属类别，并选择相应的词库，然后根据所属类别将待识词和库中词的相应特征按一定的规则进行匹配；最后利用D-S证据理论将各特征的匹配结果进行融合，选择具有最大基本概率赋值的词汇作为输出，实验结果表明，该系统不但具有较好的实时性和灵活性，而且识别率比较高。

A Recognition System of Single-Hand Words in CSL

In this paper, a recognition system of singlehand Chinese sign words based on multi-feature matching and information fusion is introduced, which uses the data glove, vision devices and sensors fixed on the elbow as input devices. At first, which category the recognized word belongs to is defined and the word library is chosen by the hand pose information and gesture information. Then the multi-feature matching is carried out in accordance with some rules. Finally, the matching results are fused by the D-S evidence theory and the word which has the largest basic probability assignment is outputted. The experiment results show that the system has not only the better real-time property and flexibility, but also higher recognition rate.