POSITION INFORMATION ESTIMATION DEVICE, METHOD THEREOF, AND PROGRAM

摘要

Observation signals xi(t) to xj(t) from a plurality of sensors arranged in two-dimensional way are subjected to the Fourier transform for a short period of time. From these, signals Xl(omega1) ~ Xl(omegaN), ~ XJ(omega1) ~ XJ(omegaN) are generated, and separation matrix W(omega1) ~ W(omegaN) are generated by the independent component analysis method. The inverse matrix of it H(omega1) ~ H(omegaN) are calculated. For an element pair of each column Hji(omegan) and Hj'i(omegan) for each omegan (n = 1, ..., N), an angle theta^i, jj'(omegan) = cos<-1> (arg(Hji(omegan)/Hj'i(omegan))/ (omegan)c<-1>||dj - dj'||)) is calculated. The arg (alpha) is a deflection angle of alpha, the c is a signal propagation speed, and the ||dj - dj'|| is an interval between the sensor j and j'. Columns are reshuffled so that the theta^i, jj'(omegan) obtained from each column is in the ascending order in the H(omega1) ~ H(omegaN). For the columns which cannot be reshuffled, ||qi - dj'|| /||qi - dj|| = ||Hji(omegan)/ Hj'i(omegan)|| = DRi, jj'(omegan) is solved for qi, and Ri, jj'(omegan) = ||DRi,jj'(omegan)(dj - dj')/(DR<2>i,jj'(omegan) - 1) || is calculated. The columns of the H(omegan) are reshuffled so that the Ri,jj'(omegan) is in the ascending order. By using this H(omegan), permutation of the W(omegan) is solved.