Gait recognition robust to dress and carrying using multi-link gravity center track

There has been increasing interest in applying gait recognition systems to real lives. Currently, the main obstacle hindering the appearance of gait recognition system is high complexity and suffering from dress and carrying variance. In this paper, it proposes a gait recognition method based on multi-link gravity center track (ML-GCT) that has low computation cost but robust to dress and carrying. It divides human silhouette into multiple links and calculates the GCT of each link. In order to improve the recognition accuracy, it takes Fisher's ratio as the importance score (IS) to rank body regions, then completes matching on the ranked regions. Test shows that the method is capable of achieving excellent recognition under different dress and carrying conditions.