基于ANN/HMM的中国手语识别系统

Sign Language Recognition Method Based on ANN/HMM

手语是聋哑人使用的语言。它是由手形动作辅之以表倩姿势为符号构成的比较稳定的表达系统，是一种靠动作／视觉交际的特殊的语言。一方面，手语识别可以作为健全人与聋哑人之间的翻译，为聋哑人提供更好的服务;另一方面，作为人体语言理解的一部分，手语识别可作为人机交互的一种手段。该文实现了基于ANN／HMM的手语识别系统，采用ANN方法建立了关于手形、位置、方向的特征映射器，并在建立手形特征映射器的过程中，给出了多特征多分类器融合算法。实验证明，基于ANN／HMM的手语识别系统是可行及实用的。

Sign language is the language used by deaf-mute, which is a steadier expressive system composed of posture motion assisted by expression pose. And it is the special language communicated by motion/vision. On one hand, sign language recognizer can be used as a translator between healthy person and deaf-mute, which provides better service for deaf mute; on the other hand, as a part of body langUage understanding, sign language recognition can be used as a method for human-computer interaction. In this paper an ANN/HMM-based sign language recognition system is described. Using ANN, feature-mapper on posture, position and orientation is built respectively. In the process of building feature-mapper on posture, multi-feature and multi-classifier fusion algorithm is presented. By experiment, it is shown that ANN/HMM-based sign language recognition system is feasible and effective.