Abstract: A gait recognition algorithm by adaptive features selecting is proposed in this paper.A novel method of gait feature extraction is proposed，which divides human body area into several sub-areas according to body unique feature and describes gait features utilizing the distance and angle of the sub-areas centroid and head centroid.Boosting method is used for adaptive selection most superior characteristic sequence and then the recognition result is processed by weighting.The proposed method combines static and dynamic gait information and the experiment result show that the approach is effective and has encouraging recognition performance.