摘要
1 引言手语是聋哑人使用的语言。它是由手形动作辅之以表情姿势而构成的比较稳定的表达系统,是一种靠动作/视觉进行交际的特殊语言。手语识别系统与手语合成系统,共同构成“人-机手语翻译器”,为聋哑人提供更好的服务。人类交互往往声情并茂,除了采用自然语言(口语、书面语言)外,人体语言(表情、体势、手势)
Sign Language is the language used by deaf-mute. In this paper the ANN/HMM-based sign language recognition method is proposed. Feature-mapper on posture , position and orientation is built respectively using ANN. The feature vector composed of output from each feature-mapper is used as input of HMM. In addition, in the process of building feature-mapper on posture, multi-feature multi-classifier fusion algorithm is presented. It is proved by experiment that ANN/HMM-based sign language recognition method is feasible and effective.
出处
《计算机科学》
CSCD
北大核心
1999年第10期63-66,共4页
Computer Science
基金
国家863计划(合同号:863-306-03-01-1)
国家自然科学基金重点课题(批准号:69789301)
国家教委跨世纪人才基金
中国科学院百人计划的资助
关键词
手语识别
参数估计
模式识别
计算机
sign language recognition, Artificial neural network, Hidden markov model, Feature-map-per
