The spotting and recognition of the human gestures is a key task in automating the analysis of the video material and human-robot interaction. Specially, applying this technology to low-resolution video has many potential applications. The human area is small with respect to input video frames in broadcast sports video, surveillance video, etc. However, this condition makes the spotting certain gesture in a video sequence a challenging task, especially if there is large camera motion. To overcome the problems, we propose a posture matching method based on curvature scale space templates of the human silhouette. We also propose a new recognition method which is robust to noisy sequences of data.