DOI Number:10.19651/j.cnki.emt.2312748

Journal:Electronic Measurement Technology

Key Words:misinginformation;cluteredbackgrounds;gesturerecognition;deplearning;contrastivelearning

Abstract:Aiming at the problem of the information missing gesture recognition based on deep learning needs a large amount of labelling. The deeper the network, the more parameters. We first collect a dataset called IMG_NUIST, consisting of missing and full gesture information. Then, we propose a new gesture recognition model.
CLGR, the inter-clas and intra-class similarities constraints enhance the feature learning performance of the model. Extensive experiments are conducted on two classic datasets (ASL Alphabet and NUSI) and the proposed IMG_ NUIST dataset. The experimental results are shown as follows: 1)in the ablation study, contrastive learning can effectively % and the model convergence speed is significantly accelerated. 2)In the comparative experiments with two recent works and two contrastive learning models, the computational complexity of CLGR is 41.4% simpler than that of the two comparison models on average.CLCR can recognize the gestures with missing information and works well for those gestures with cluttered backgrounds. The gesture recognition accuracy of CLCR on the NUSI and IMG_ NUSIT datasets outperforms the four comparison methods and is only 0.43% lower than the best result on ASL. Especially on the NUSI dataset, CLCR increases the recognition accuracy of gestures by 17.35% on average. The experimental results show that the proposed models are significantly effective for gesture recognition tasks with missing information and cluttered backgrounds with fast convergence speed and low computational complexity, and it is practical.

Indexed by:Journal paper

Volume:46

Issue:7

Translation or Not:no

Date of Publication:2023-07-01