Abstract:

With the emergence of RGB-D sensor, depth image information can make up for the poor robustness of human body gesture recognition based on color image under complex environment and light changes. Therefore, it becomes more convenient to use depth information to identify human body gesture. To understand the current development situation of human body gesture recognition, the extensive research is expounded on the basis of existing literature and the latest results from the depth image preprocessing, feature extraction and gesture recognition algorithm. The technical development and the existing difficulties and problems are discussed for future related research.

Key words: body gesture recognition, RGB-D, posture classification, depth information

摘要：

随着RGB-D传感器的出现，深度图像信息可以弥补基于彩色图像的人体姿态识别在复杂环境和光照变化下鲁棒性较差的问题，因此利用深度信息识别人体姿态变得更加便捷。为了解人体姿态识别的发展现状，在广泛调研现有文献和最新成果的基础上，从深度图像预处理、特征提取、姿态识别算法三方面对基于深度信息的人体姿态识别进行阐述，介绍人体姿态识别相关的技术发展及应用领域，并对其中存在的难点与问题进行讨论，为以后的相关研究提供思路。

关键词: 人体姿态识别, RGB-D, 姿态分类, 深度信息