Abstract: A 3D facial motion tracking approach is proposed based on the incorporation of online appearance model (OAM) and cylinder head model (CHM) in framework of particle filter. It includes: 1) For the construction of OAM, multi-measurements are infused to reduce the influence of lighting and person dependence. 2) OAM can provide the detailed descriptive parameters used for facial motion tracking, however, it is not suitable for robust facial motion tracking across large pose variation. To alleviate these problems, OAM and CHM which is suitable for robust global head motion tracking across large pose variation, are combined, where the global head motion parameters obtained from the CHM are used as the cues of the OAM parameters for a good fitting. And the good fitting result is set as the initial of CHM in next frame. 3) Motion filtering is applied by particle filter combined with local optimization and improved resampling. Experiments of tracking real video sequences demonstrate that accurate tracking is obtained even in the presence of perturbing factors including significant head pose and facial expression variations, occlusions, and illumination changes. And it is also shown that facial motion tracking combining OAM and CHM is more pose robust than that of OAM in terms of 24% higher tracking rate and 11% wider pose coverage. The between subjective experiment indicates the suitability of subjective face identification on synthesized video with tracked facial motion parameters by it.