基于特征聚类表示学习的静态手势识别
首发时间:2021-03-25
摘要:人体手势含有丰富的语义信息,是当前人机交互领域的应用热点。静态手势动作受到动作者手型差异和主观动作程度的差别影响,导致使用深度网络训练识别模型时特征与标签对应性差,手势类别间边界不清,最终导致识别模型准确率低。本文提出一种基于特征聚类评价的静态手势特征表示学习识别方法,以解决手势特征对应关系较差引起的特征分布边界不清的问题,提高网络提取特征的聚合度,从而提高手势识别准确率。通过对公共数据集的测试,并与改进前的方法进行对比分析,从而证明本文所提出方法的有效性。
For information in English, please click here
Static gesture recognition based on feature clustering representation learning
Abstract:Human gestures contain rich semantic information and are currently a hot application in the field of human-computer interaction. Static gestures are affected by the difference of the actor\'s hand type and the degree of subjective action, which leads to poor correspondence between features and labels when using deep network to train the recognition model, and the boundaries between gesture categories are unclear, which ultimately leads to low accuracy of the recognition model. This paper proposes a static gesture feature representation learning method based on clustering feature evaluation to solve the problem of unclear feature distribution boundaries caused by poor correspondence between gesture features, improve the aggregation degree of network extracted features, and improve the accuracy of gesture recognition. The effectiveness of the method proposed in this paper is proved by the test on the public data set and the comparative analysis with the method before the improvement.
Keywords: deep learning static gesture feature clustering degree of clustering
基金:
引用
No.****
动态公开评议
共计0人参与
勘误表
基于特征聚类表示学习的静态手势识别
评论
全部评论0/1000