一种基于证据理论的中国手语单手静态词识别方法

单手静态词是中国手语词汇库的一个重要组成部分,该类词主要借助于单手的固定姿态和手同身体之间的相对位置关系进行表达。结合单手静态词的特点,提出了一种基于D-S证据理论的中国手语单手静态词汇识别方法:利用数据手套CAS_Glove获取手姿信息,借助于视觉和肘部弯曲传感器融合确定出手在人体空间中的相对位置;在对单手词进行模糊特征空间划分的基础上,利用D-S证据理论实现了多种特征匹配结果的融合。为减少计算复杂度,增强系统实时性,对Dempster规则在特定条件下进行了简化,并从理论的角度对其进行了证明。最后通过实验比较证实了该方法的有效性。

Recognition Method of Single-hand Static Sign Words in CSL Based on Evidence Theory

Static singe-hand sign word is an important component in Chinese sign library, which is mainly expressed by hand pose and the relative position between hand and body. With characteristics of this kind word, a recognition method based on the D-S evidence theory was introduced: hand pose information was captured by the dataglove CAS_Glove and the position information between hand and body was calculated out by fusing sensing of vision and elbow sensor; after partitioning of fuzzy feature space, the multi-feature matching results were fused by D-S evidence theory. In order to reduce the calculation complexity and enhance the real-time property, the Dempster rule was simplified under appropriate conditions, and the rationality of this simplification was demonstrated in theory. In experiments, the validity of this method was confirmed comparatively.