Clustering algorithm for static gait recognition based on low-dimensional plantar pressure features

• Published in: Ji suan ji ying yong yan jiu Vol. 32; no. 7; pp. 2176 - 2183
• Main Authors: Fang, Zhengwen, Wang, Nian, Jiang, Jinjian, Bao, Wenxia
• Format: Journal Article
• Language: Chinese
• Published: 01.07.2015
• Subjects: Accuracy; Algorithms; Categories; Clustering; Feature extraction; Fuzzy; Plantar pressure; Reduction
• DOI: 10.3969/j.issn.1001-3695.2015.07.063
• ISSN: 1001-3695

This paper proposed a static gait clustering algorithm to explore the application of plantar pressure features in personal identification. Firstly, the algorithm extracted the common plantar pressure features including global and local ones from the static data through a pressure platform, which could be a vector to characterize the samples. Then it used non-negative matrix factorization(NMF) method to get the mapping of the samples in the transformed feature space and the representation in low-dimensional space. Finally, after dimensionality reduction, these features were to be clustered by fuzzy C-means algorithm(FCM). The result shows that the clustering accuracy can reach about 90% and compared with other methods, it has advantages in accuracy at the same time. Through the experiment and comparative analysis, the method which compressed the sample data to a very low-dimensional feature space can retain the category information of samples, and then draw the conclusion that the plantar pressure features ext