听障教学中手语识别技术的研究进展

Research Progress of Sign Language Recognition Technology in Hearing Impaired Teaching

手语是听障生课堂教学中的主要交流手段,但以健听人为主的手语识别越来越不能满足听障生课堂教学的需求,因此,研究手语识别技术辅助听障教学是非常必要的。文章对国内外手语识别的纯技术研究做出梳理。对传统的手语识别方法进行归纳,包括模板匹配方法、隐马尔可夫模型、神经网络及多种方法融合的手语识别技术。重点研究基于深度学习的手语识别技术,包括基于卷积神经网、循环神经网络、图神经网络及多种方法融合的手语识别技术。

Sign language is the main means of communication in classroom teaching for hearing impaired students,but sign language recognition based on healthy listeners is increasingly unable to meet the needs of classroom teaching for hearing impaired students.Therefore,it is very necessary to study sign language recognition technology to assist the teaching of hearing impaired students.This paper reviews the pure technical research of sign language recognition at home and abroad.This paper summarizes the traditional sign language recognition methods,including template matching method,hidden Markov model,neural network and multimethod fusion of sign language recognition technology.It focuses on sign language recognition technology based on deep learning,including sign language recognition technology based on convolutional neural network,recurrent neural network,graph neural network and the fusion of various methods.