摘要
Face recognition from video requires dealing with uncertainty both in tracking and recognition. This paper proposed an effective method for face recognition from video. In order to realize simultaneous tracking and recognition, fisherface-based recognition is combined with tracking into one model. This model is then embedded into particle filter to perform face recognition from video. In order to improve the robustness of tracking, an expectation maximization (EM) algorithm was adopted to update the appearance model. The experimental results show that the proposed method can perform well in tracking and recognition even in poor conditions such as occlusion and remarkable change in lighting.
Face recognition from video requires dealing with uncertainty both intracking and recognition. This paper proposed an effective method for face recognition from video.In order to realize simultaneous tracking and recognition, fisherface-based recognition is combinedwith tracking into one model. This model is then embedded into particle filter to perform facerecognition from video. In order to improve the robustness of tracking, an expectation maximization(EM) algorithm was adopted to update the appearance model. The experimental results show that theproposed method can perform well in tracking and recognition even in poor conditions such asocclusion and remarkable change in lighting.
基金
National Natural Science Foundation of Chi-na(No.60375008)
Shanghai Key Technolo-gies Pre-research Project(No.035115009)
