Non-specific Human Sign Language Recognition Based On Deep Learning

Sign language recognition is an important research direction in the field of computer vision.It is a multidisciplinary research topic that combines image processing,pattern recognition and machine translation.The social and practical significance of sign language recognition has led more and more researchers to develop effective sign language recognition systems to help hearing-impaired people better integrate into society,enjoy social public resources,and experience the many conveniences that science and technology bring to people's lives.At present,sign language recognition researchers use traditional machine learning methods to carry out relevant research on the topic,and have achieved good academic research results in small-scale sign language databases.The drawbacks of traditional machine learning models are revealed.With the fierce rise of deep learning,the network modeling method of neural networks has brought breakthrough progress to various research fields,and also provided new research ideas for sign language recognition.Therefore,this paper designs a sign language recognition framework based on the deep learning network model.The main research contents are as follows:First of all,the research significance and current situation of sign language recognition are expounded,and some research achievements and research progress of sign language recognition are summarized and analyzed.It also summarizes the problems that have not yet been solved in the study of sign language recognition,and explores a model that is more suitable for the task of sign language recognition research.Then,Introduced the study of sign language recognition based on traditional machine learning algorithms in detail,and summarized the principles and advantages of feature algorithms and classifiers involved in mainstream sign language recognition strategies.The shortcomings and problems of mainstream methods are analyzed in combination with the application examples of sign language recognition,and the internal reasons for the limitations are analyzed in depth.Furthermore,aiming at the feature design problem of static sign language gesture word recognition task.This paper builds an end-to-end sign language recognition system based on deep residual neural network combined with data enhancement technology,compared the traditional machine learning strategy based on SIFT-SVM and shallow convolutional neural network model,comparison of recognition performance on ASL dataset,verified the robustness and high accuracy of this method.Finally,aiming at the problem of sequence modeling in the research on the recognition of isolated words in sign language based on video,a sign language recognition model based on long-short-term-memory networks and connectionist temporal classifier.The model uses a convolutional neural network as a feature extractor,obtaining mid-level features as features of sign language video samples.Build a classifier model based on BiLSTM-CTC for isolated sign language recognition,through experiments to prove the effectiveness of the method.