基于Kinect传感器的人体行为分析算法

Human behavior analysis algorithm based on Kinect sensor

人体行为分析一直是计算机视觉领域中具有挑战性的研究方向,近年来深度传感器的引入为解决人体行为分析问题提供了新的研究方法。采用微软Kinect传感器获取深度图像,首先对深度图进行局部梯度特征提取,再结合条件随机场(CRF)模型,提出一种新的人体行为分析方法。实现了对简单人体行为的有效识别,通过在2个流行人体行为数据库上实验,证明了该方法具有较好的识别结果和该方法的优越性能。

Human behavior analysis has been a challenging research in the field of computer vision, in recent years, to solve the problem of human behavior analysis, introduction of depth sensor provide a new research method. Depth map captured by Microsoft Kinect sensor, firstly,local gradient features extraction of depth map are carried out, in addition, couple^with conditional random field model（ CRF）, put forward a new approach of human behavior analysis. This method is used to effectively recognize several simple human behavior. Experiments are tested on two popular human behavior databases, it verifies that this approach has good recognition result and achieves superior performance.