Sign language recognition method based on deep learning

Abstract

The invention discloses a sign language recognition method based on deep learning. The steps are: (1) divide the database sample set; (2) collect image blocks, (3) whiten the data; (4) train the sparse autoencoder network; (5) obtain the convolution feature map; (6) obtain the pooling feature Figure; (7) Training classifier; (8) Test classification results. The present invention uses the reverse conduction algorithm to train the sparse self-encoding network, so that the present invention improves the recognition rate when processing complex background data. The present invention selects the weight of the sparse self-encoding network as the convolution kernel, and obtains the convolution by convolution The feature map combines supervised learning and unsupervised learning to reduce the manpower and material resources of manual labeling. The present invention adopts the maximum pooling method to obtain pooled feature maps, reduce the feature dimension, and reduce the time required for sign language recognition tasks. Complexity.

Description

technical field [0001] The invention belongs to the technical field of image processing, and further relates to a sign language recognition method based on deep learning in the technical field of pattern recognition. The invention can be used for human body sign language recognition based on syllable syllable simulation according to gesture changes and sign language information exchange between people and computers. Background technique [0002] The research of human-computer interaction technology is an important part of the field of computer technology research. When people communicate face-to-face, it includes natural language such as spoken language and written language, as well as body language such as sign language, facial expressions, body gestures and mouth shapes to transmit information, so research on the perception model of human body language and its information fusion with natural language , which is of great significance for improving the level of computer nat...

AI Technical Summary

Problems solved by technology

The features extracted by the above methods can be identified by classifiers such as support vector machines, but the shortcomings of these methods are that the extracted features are relatively low-level, and the extracted features are greatly affected by the complex background, and the implementation is complicated, which leads to this method in processing When there is a large amount of complex background data, the accuracy of sign language recognition is reduced due to the lack of robustness

The shortcomings of this method are that the image data background requirements are relatively simple, the difference between sign language categories is relatively large, and the accuracy is relatively low when dealing with complex data background and complex sign language recognition tasks. way, increasing the manpower and material resources of manual labeling

The disadvantage of this method is that the process of calculating the Euclidean distance between the image to be tested and the standard image feature vector to achieve classification is too computationally intensive and the computational efficiency is too low, so it is not suitable for large-scale data processing

Images