一种单手动态手指语的识别方法

本文利用数据手套CAS-Glove作为输入设备，在将各手指关节的运动状态进行模 糊划分的基础上，提出了一种基于多特征匹配和D-S证据理论的单手动态手指语识别方法． 该方法以初始状态向量、状态概率分布和周期特征向量作为词汇的特征描述，通过特征匹配 确定候选词的基本概率赋值，根据D-S证据理论将这些信息进行融合，确定出最终的识别结 果．实验结果表明该方法具有比较好的识别率和实时性，证实了该方法的有效性．

A RECOGNITION METHOD OF SINGLE-HAND DYNAMIC FINGER WORDS IN CSL

In this paper, a novel recognition method of single-hand dynamic finger words of Chinese Sign Language (CSL) based on multi-feature matching and D-S theory is introduced, which uses the dataglove CAS-Glove as the input device and partitions the motion of each finger joints into a fuzzy state. In this method, each word is represented by the vector of initial appearance, probability distribution of each state and the vector of the periodic property. The basic probability assignments of candidate words are obtained by feature match and fused to achieve the best words by D-S theory. The results of experiments show that this method has higher recognition rate and better real-time property, so the effectivity of this method is confirmed.