Face recognition for images from the same unknown person
Yea‐Shuan Huang, Yao-Hong Tsai, Hong-Hsin Chao, Yi-Tsung Chien · 2004
This paper mainly introduces (1) a face recognition method by using a newly designed radial basis function (RBF) neural net which can iteratively reduce a purposely defined classification-oriented error function, and (2) a decision-making mechanism by accumulating multiple individual face recognition results of the same unknown targeted person. To experiment on 50 persons (each person has 32 training samples and 100 testing samples), although the recognition rate of individual sample is only 86.5%, a perfect recognition accuracy (i.e. 100% accuracy) has been achieved by accumulating 20 temporal face images. This shows that the proposed approaches can produce various degrees of security to support different face recognition applications.