Proper orthogonal projection - multiple signal classification (POP-MUSIC)

Proper Orthogonal Projection - Multiple signal classification (POP-MUSIC) is introduced. Two spectral estimates are derived using geometrical approach based on POP-MUSIC which appear to be the counterparts of the Maximum Likelihood (ML) spectral estimate and Linear Prediction (LP) spectral estimate. POP-MUSIC based on linear prediction (POP-MUSIC-LP) provides a superior resolution compared to POP-MUSIC based on maximum likelihood (POP-MUSIC-ML), and this is justified mathematically. Quantitative measures such as degrading factor due to spurious peaks (F) and sharpness factor (Y) are introduced to facilitate the comparative performance of the spectral estimates. A systolic array structure which is suitable for VLSI implementation is given for adaptive estimation of the cross spectral density matrix and the POP-MUSIC spectral estimation. Computer simulation is presented.