Sign language recognition method based on convolutional neural network

Abstract

The invention discloses a sign language recognition method based on a convolutional neural network. The sign language recognition method comprises the specific steps that 1, a plurality of depth images containing sign languages are collected; 2, segmenting the hand shape parts in all the depth images from the background through a preprocessing step to obtain a complete noise-free hand shape image,and establishing a sign language image database; dividing the hand-shaped image in the sign language image database into two parts, one part serving as a training sample, and the other part serving as a test sample; constructing a convolutional neural network model; 3, training a convolutional neural network model pair by using the training sample; and 4, identifying the test sample by using thetrained convolutional neural network model, and outputting a classification and identification result. According to the sign language recognition method based on the convolutional neural network, theproblem that a traditional sign language recognition method is low in efficiency for tasks with large data volumes is solved.

Description

technical field [0001] The invention belongs to the technical field of sign language recognition, and relates to a sign language recognition method based on a convolutional neural network. Background technique [0002] Sign language is the only way for hearing-impaired people to communicate with normal people. It is precisely because of the sign language recognition system that the living and working space of the deaf-mute people is not restricted. On the other hand, with the development of artificial intelligence technology, sign language, as a new and more convenient way of human-computer interaction, has become a new trend in the way of interaction in various industries today. [0003] Traditional sign language recognition methods mainly involve two steps: feature extraction and learning recognition. Artificially designed features mainly include histogram of gradient orientation (HOG) and histogram of optical flow orientation (HOF). Traditional models and methods for ti...

AI Technical Summary

Problems solved by technology

[0004] The purpose of the present invention is to provide a sign language recognition method based on convolutional neural network, which avoids the problem of low efficiency of traditional sign language recognition methods for tasks with a large amount of data

Images