Research On Signer-independent Sign Language Recognition Based On Improved HMM And Adaptive Technology

Sign language is a language used by deaf and mute for communication.The world a total of about 230 million deaf people,as the social vulnerable groups of deaf people often encounter life,work,psychology,education,communication and many other issues.Sign language of the popularity of poor,able to master the number of sign language is very small,which makes deaf and mute communication there is a big obstacle.Sign language recognition research since the nineties of last century,It through a certain computer technology to sign language can be understood in the form of ordinary people,so as to help deaf and mute to express and communicate.This technology,as one of the research contents in the field of human-computer interaction,has important practical significance to the deaf-mute group.At present,the accuracy rate of sign language recognition based on data gloves has reached a good level,and the recognition of signer-dependent sign language recognition has better performance,but the performance of signer-independent sign language recognition system is relatively low.And the signer-independent sign language recognition is the key problem to be solved urgently when the sign language recognition system is applied.The difference of individual sign language data and the lack of sign language training samples are important reasons for the performance of signer-independent sign language recognition system.Aiming at this problem,this paper studies the key algorithms in sign language recognition,and improves the existing algorithms.The main work of this paper is as follows:(1)The training model based on Hidden Markov Model(HMM)is studied and analyzed.The Guided Adaptive Evolutionary Genetic Algorithm is used to train the HMM parameters of sign language,so as to find the global optimal solution in a given solution space,which improves the problem that the Baum-Welch algorithm converges to the local optimal solution and improves the accuracy of sign language recognition.(2)An adaptive sign language recognition framework based on the Maximum Likelihood Linear Regression(MLLR)algorithm and the Maximum A Posteriority(MAP)algorithm is proposed.This approach optimizes the division MLLR regression class to provide more accurate initial MAP model,which give full play to the rapidity and the MAP MLLR progressive.Then introduced MCE model parameter estimation algorithm to compensate for the limitations of the model parameters adaptive method to further reduce the system error rate and accelerate the recognition speed.Meanwhile,for the MCE algorithm computationally intensive problems proposed improvements.The experimental results show that the performance of this algorithm is better than that of the existing adaptive algorithm in signer-independent sign language recognition.The algorithm can make a good effect of signer-independent sign language recognition by using a small amount of adaptive sign language data.