Background extraction device, principal component analysis device and principal component analysis program

摘要

PROBLEM TO BE SOLVED: To reduce a calculation load required for a main component analysis.SOLUTION: There is provided a main component analysis device which calculates eigenvectors u(N),..., u(N) with respect to N P-dimensional vectors. Covariance matrices Cfor a part of L vectors are derived, and eigenvectors u(L),..., u(L) are calculated. With respect to k=L+1,..., N, the following processing is repeated: on the basis of a vector V, a unit vector orthogonal to eigenvectors u(k-1),..., u(k-1) is calculated and set as an eigenvector u(k-1); an M+1-dimensional covariance matrix C' is approximately derived as a covariance matrix Cof vectors V,..., Vby using eigenvector u(k-1),..., u(k-1); and an eigenvector of the covariance matrix C' is calculated and converted into P-dimensional eigenvectors u(k),..., u(k).