Classification Accuracy of Personal Identification Based on Joint Motions Using 2D Information

Ryusuke Miyamoto
2018 Journal of Computers  
This paper evaluates the classification accuracy of personal identification by a classification scheme with feature extraction based on joint motions using only two-dimensional information. Experimental results show that the feature extraction based on joint motions can achieve moderate classification accuracy when feature vectors are constructed from only two-dimensional information in an image plane. In addition, the results include interesting knowledge: the classification accuracy is not
more » ... accuracy is not degraded drastically even if a gait is measured from right in front. In the best case, the classification accuracy becomes 78.95% in the experiment and it is 75.44% in the worst case.
doi:10.17706/jcp.13.1.49-57 fatcat:kgox37ej2zcmfe2onde46lfjqq