基于深度信息的人体动作识别研究综述

A review for human action recognition based on depth data

随着低成本深度传感器的发明,尤其是微软Kinect的出现,高分辨率的深度与视觉(RGB)感知数据被广泛使用,并为解决计算机视觉领域中的基本问题开拓了新的机遇。本文针对基于深度信息的人体动作识别研究,首先提出了一种基于特征和数据类型的分类框架,并对最近几年提出的相关方法进行了全面回顾。随后,对文献中描述的算法进行了性能对比分析,同时对所引用的公共测试数据集进行了总结。最后,笔者对未来的研究方向进行了讨论并给出了相关建议。

With the invention of the low-cost depth sensors,especially the emergence of Microsoft Kinect,high-resolution depth and visual （RGB）sensing data has become available for widespread use,which opens up new opportunities to solve fundamental problems in computer vision commu-nity.This paper presents a comprehensive review of recent depth-based human action recognition algorithms.Firstly,we develop a taxonomic framework according to features and original data type.Following our taxonomy,recent published research on the use of depth data for recognizing human action is reviewed.Then,the publicly available datasets cited in their work are listed.Fi-nally,the authors discuss and suggest future research directions.