Sign language recognition model based on deep learning method

Abstract

The invention relates to a sign language recognition model based on a deep learning method. The sign language recognition model comprises a picture reading module, a data enhancement module, a skin color segmentation module, a feature extraction module, a training model module and a comparison output module. The system is characterized in that the picture reading module is completed by using an OpenCV visual library; the data enhancement module adopts dropout in deep learning to effectively transform an original image; the skin color segmentation module carries out skin color segmentation by adopting an Otsu threshold value method; the feature extraction module obtains a sampling feature graph by using a Focus structure; the training model module adjusts parameters of a YOLOV5 target detection model according to the processing effect to obtain a weight file; and the comparison output module finds a picture feature prediction logic maximum value by using a prediction function pre-dict, and outputs a text recognition result. The sign language recognition method has the advantages of being high in accuracy, high in light interference resistance and the like, the recognition rate of sign language recognition obtained through debugging is 94.57%, and interference of complex backgrounds on sign language recognition can be reduced to a certain degree.

Description

technical field [0001] The invention relates to a sign language recognition model based on a deep learning method, which is mainly used in the technical field of sign language translation. Background technique [0002] There are millions of hearing-impaired people around the world who usually use sign language to communicate. There are more than 20 million people with hearing and speech disabilities in my country, and the number is increasing at a rate of 20,000 to 30,000 every year. However, communication barriers between deaf-mute and able-bodied people are not uncommon, and deaf-mute people always have the problem of being difficult to integrate into society. Sign language recognition is a problem that has been solved in many years of research. In the early days, it was mainly based on gloves to locate the hand, identify the position of the hand joints, and judge the meaning it represents, or use the sign language translation APP to identify special points. However, this ...

AI Technical Summary

Problems solved by technology

[0004] The purpose of the present invention is to solve the problem that the traditional static gesture recognition method has low recognition accuracy and incomplete recognition of gestures in complex backgrounds. A sign language recognition model based on deep learning methods is designed, based on the YOLOV5 model and the pytorch framework. Basic, use pycharm software to write and debug the program; the model can basically accurately segment the skin color area of ​​the human body for photos under normal light intensity. The algorithm has accurate segmentation effect and good robustness. Finally, it is proved by experiments The accuracy rate of gesture recognition reaches 94.57%

Images