| 摘要 |
<P>PROBLEM TO BE SOLVED: To decide a face image with a small amount of data. <P>SOLUTION: In this information processor, because an image of a label W11 becomes an image classified into a label W12 when roll-rotating the image of the label W11 by +20 degrees, two pixel positions obtained when learning the image of the label W12 can be set as positions wherein positions of two pixels obtained by learning the image of the label W11 are roll-rotated by +20 degrees. That is, when evaluation value calculation processing to a non-learning node 81-1-1-2 or a non-learning node 81-1-1-2-1 is performed, a pixel difference characteristic d of the pixels of the positions wherein the positions of the two pixels are roll-rotated by +20 degrees used so as to calculate a weak hypothesis f(x) by a weak discriminator of an already learned node 81-1-1-1 or an already learned node 81-1-1-1-1 learning the label W11 on a window image is calculated. The information processor can be applied to identification of a face of a person in a robot. <P>COPYRIGHT: (C)2007,JPO&INPIT |
