结合HMM和SPSM的手语识别方法  

Combining HMM and SPSM for sign language recognition

作　　者：周宇[1] 陈熙霖[2] 王春立[3] 赵德斌[1] 高文[1] 

机构地区：[1]哈尔滨工业大学计算机科学与技术学院,哈尔滨150001 [2]中国科学院计算技术研究所,北京100190 [3]大连海事大学计算机科学与技术学院,辽宁大连116026

出　　处：《计算机工程与应用》2008年第21期37-40,共4页Computer Engineering and Applications

基　　金：国家自然科学基金(the National Natural Science Foundation of China under Grant No.60603023, No.60533030);北京市自然科学基金(the Natural Science Foundation of Beijing City of China under Grant No.4061001)

摘　　要：手语识别的研究具有重大的学术价值和广泛的应用前景。在近些年的手语识别工作中,隐马尔可夫模型(Hidden Markov Models,简称HMMs)起到了重要的作用,但是,HMMs假设同一状态内的观察值之间是独立同分布的,这个假设同某些手语信号的帧间相关性相背离。受到多项式片段模型(Polynomial Segment Models,简称PSMs)能够显式描述帧间相关性的启发,提出了一种简化的PSMs,其中应用马氏距离作为距离测度。实验表明,这种简化的PSMs在同传统的HMMs进行后验概率归一化求和的融合之后,手语词的平均相对正确率得到了13.38%的提升,从而证明此方法是一种更加精确的手语识别方法。The research of sign language recognition has great academic value and broad application prospect.In recent works on sign language recognition,Hidden Markov Models（HMMs） has played an important role.But the framework of HMMs makes an assumption that the observations in the same state are independent and identically distributed,which is not consistent with sign language signals sometimes.Inspired by Polynomial Segment Models（PSMs） that can model the consecutive frame correlation well, in this paper we propose a type of simplified PSMs,in which Mahalanobis distance is used as the similarity metric.Experimental results show that after combining conventional HMMs and the simplified PSMs with summation of normalized posteriori,the average relative accuracy can be improved by 13.38%.As result,our method is superior to conventional HMMs.

关 键 词：手语识别 人机交互 隐马尔可夫模型 多项式片段模型 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]