基于高斯过程的表情动作单元跟踪技术  

Facial Tracking by Gaussian Process

作　　者：王磊 邹北骥[2] 彭小宁[2] 潘丽丽[1] 

机构地区：[1]湖南大学计算机与通信学院 [2]中南大学信息科学与工程学院

出　　处：《电子学报》2007年第11期2087-2091,共5页Acta Electronica Sinica

基　　金：国家自然科学基金(No.60573079;No.60673093)

摘　　要：在表情动作单元的跟踪中有两个常见问题:一是跟踪结果有小幅而频繁的抖动;二是跟踪过程会产生难以检测的误差.针对这两个问题,本文提出了一种基于高斯过程和粒子滤波的表情动作单元跟踪技术.实验结果表明本文算法比传统的梯度优化和粒子滤波法具有更好的平滑性和跟踪精度,而精度的优势在头部有偏转的情况下尤为突出.Facial Action Units（FAU）tracking is a hard problem for the rigid and non-rigid transformations of human face. The constantly trembling in the tracking result and the tracking failures caused by the absence of constraint remain open problems. This paper presents a novel method to attack these problems by combining Gaussian Process and Particle Filtering.Gradient-based method and particle filtering based method are compared with our method and the experiment results are encouraging.

关 键 词：表情动作单元跟踪 梯度优化 粒子滤波 高斯过程 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]