Manual alphabet identification method based on RGB-D image

Abstract

The invention discloses a manual alphabet identification method based on an RGB-D image, and belongs to the technical field of behavior identification of computer vision. According to different types of hand texture information of different types of manual alphabets, histogram-of -oriented-gradient characteristics and super-normal-vector characteristics of an RGB-D image video frame obtained through an RGB-D camera are extracted; a principal component analysis and a linear discrimination analysis are combined for characteristic attribute optimizing processing, so that a characteristic attribute with saliency is obtained, and BoW sparse representation is conducted on the characteristics; a data contrast mining technology is used for obtaining representative exemplars in all the types of manual alphabet; finally, a hierarchy judgment policy based on a greedy thought is adopted, the manual alphabets easy to classify are quickly classified through a nonparametric k-Nearest Neighbor Algorithm classifier firstly, and the manual alphabets difficult to classify are judged through a trained support vector machine model based on templates. Compared with the prior art, the manual alphabet identification method is high in both identification accuracy and identification efficiency.

Description

technical field [0001] The invention relates to a sign language letter recognition method, in particular to a sign language letter recognition method based on an RGB-D image, and belongs to the technical field of computer vision behavior recognition. Background technique [0002] Deaf-mute people account for a large part of human society, but it is difficult for deaf-mute people to integrate into social life. This is because the deaf-mute use sign language as the main way of communication, but the reality is that most ordinary people cannot understand sign language, which hinders the integration of the deaf-mute into society. The purpose of studying sign language recognition is to establish a communication bridge between the deaf-mute and ordinary people, and realize barrier-free information exchange between the deaf-mute and ordinary people. Sign language is a human language that integrates a variety of human communication methods including gestures, expressions, and lip m...

AI Technical Summary

Problems solved by technology

[0007] Some researchers have proposed some methods for sign language recognition using RGB-D data. For example, a Chinese invention patent application discloses a "Gesture recognition method based on a small number of training samples composed of RGB-D data" (application date is 2013 -10-29, the publication date is 2014-01-22, and the application number is 201310522370.7), this method can use less RGB-D data samples to realize gesture recognition, but the patent uses complex optical flow tracking algorithm to detect human motion points of interest, the computational complexity is high; in addition, the random distribution of non-zero coefficients in the sparse feature vector generated by the traditional sparse coding method used in this patent reduces the inter-class separation characteristics of sparse features; finally, the patent uses the recent The final result is judged by the neighbor classifier. Although the classifier is simple and easy to use, it is prone to misjudgment for intra-class difference samples

Another Chinese invention patent application discloses "A Kinect-Based Chinese Sign Language Recognition Method" (the application date is 2013-5-28, the publication date is 2014-9-5, and the application number is 201310204961.X). Only the PCA dimensionality reduction method is used to optimize the acquired features, but in some cases the PCA dimensionality reduction method will have a negative effect, extracting feature attributes that do not have inter-class separation characteristics, thus affecting the accuracy of the final sign language recognition results

[0008] In summary, the existing sign language recognition technology based on RGB-D data generally has the problems of high computational complexity and low recognition accuracy.

Images