F0102(1) ニューラルネットワークとファジイ推論を用いた癒し構造の構築([F0102]癒し工学とその周辺工学,先端技術フォーラム)

概要

論文の詳細を見る
We have developed the approach to obtain a membership function on a panelist's favorance by using Fuzzy Quantificaion Theory2 as the output of panelist's evaluation for an input of the data obtained from images of facial expressions. The matrix dealt in the resulting equation for eigenvalues includes positive semidefinites and repeated eigenvalues, is ill-conditioned. We found that the result obtained by holographic neural networks (HNN) corresponded to one based on transforming the generalized eigenvalue problem into standard eigenvalue problem by using Moore-Penrose inverse. It is shown that the usage based on the combination of the proposed approach with HNN will be available to the research for facial expression analysis in future.
一般社団法人日本機械学会の論文
2009-09-12