Abstarct: Nowaday, computer-baxsed security systems improve the speed and accuracy of verification. In these systems, exclusive features such as biological and behavioral features are used to verificate a person. In this thesis, we proposed a video-baxsed signature verification system. A camera which is located in front view, records the signature video. The proposed method for signature verification has three main steps. First, after detecting the signature-relevant start and stop frxames, we track the pen-tip in this interval. Morever in this step, we use Particle Filtering and K-Means algorithms consecutively to reduce tracking error. Second step performs feature extraction from signature frxames. These features are in two categories, dynamic and static. The dynamic features include hand speed profile, wrist speed profile, the sequence of distance between pen-tip and signature centroid and finally sequence of angles between two lines drawn across any three sequential pen-tips. Signature duration, major and minor hand diametet ratio and right or left-handed are the static features. In the third step, after feature extraction we design a signature verification system. In this system, a model assigned to each person is according to dynamic and static features extracted from training videos. This model contains a feature vector for each person. This vector includes dynamic and static parts. Dynamic parts obtained baxsed on DTW algorithm. The static part of the vestor iscomputed from average of static features extracted from training videos. Now for verification of a claimed signature, we compare it against thr relevant model. To implementation of the propsed method, we obtain a databaxse. The databaxse involves nineteen genuine and five feigned signature videos for each of 82 persons.We used several criterions to evaluate the proposed method. In the best condition,The proposed system achieved an equal error rate of 3.8 % and Accuarcy of 98.51 % against this databaxse