步态作为唯一具备远距离识别能力的生物测量特征已经受到广泛的关注。步态序列包含人行走的静态和动态信息，综合利用这两方面信息是提高识别性能的关键。为了综合利用人行走的静态和动态信息来提高识别能力，提出了一种用步态的不变矩傅氏级数系数的幅值作为识别特征的步态识别方法。因为不变矩描述了人运动的静态信息，其在整个步态周期提取的特征则蕴含了人运动的动态信息，所以将不变矩作为识别特征用于步态识别。该方法首先计算每帧图像的不变矩；然后采用傅里叶级数来拟合整个不变矩系数序列，并用遗传算法搜索傅里叶级数系数；接着将这些系数的幅值表示为用于分类的特征向量；最后再用k近邻分类器对特征向量进行分类。通过对CMU步态数据库中的4种步态分别进行的实验结果表明，该方法对单独的矩可取得80%以上的识别率，而对级联的矩识别率则可达到90%以上。另外，该方法对部分遮挡也具有鲁棒性。实验结果和性能分析表明，这种结合静态和动态信息的识别方法是有效的。

Gait as a biometric with the unique capability to recognize people at a distance is subject to increasing interest. A gait sequence contains static and dynamic components from the walking way. It is pivotal to integrate them to improve the performance of gait recognition. Initiated from the idea of integration, a moment invariants based scheme for gait recognition is proposed in the paper, taking the magnitudes of the Fourier series coefficients representing moment invariants of gaits as features for identification. The moment invariants describe the static components during the walk, whereas dynamic components are contained in the coefficients extracted according to the whole gait sequence. So firstly, the moment invariants of each frame are computed. Secondly, the moment invariants of humans silhouettes are represented with Fourier series, the Fourier coefficients of which are obtained using a genetic algorithm. Thirdly, the magnitudes of the coefficients are generated as vectors to classify the subjects, which are identified by the kNN classifier. The recognition results of four kinds of gaits in the CMU gait database show that the proposed scheme has a correct recognition rate of more than 80% using a single moment and beyond 90% using jointed moments. Moreover, the scheme is also robust to partial occlusion. The experimental results and performance analysis indicate that the scheme is effective as it integrates static and dynamic components for identification.