Abstract:

For non-stationary array signal processing, time-frequency analysis will help to separate sources and improve the precision of direction of arrival (DOA) estimation. An adaptive local polynomial Fourier transform (LPFT) method is proposed to improve the energy concentration of time-frequency representation for multi-component signals. The window width and coefficients of LPFT are determined by performing polynomial curve fitting on the signals’ instantaneous frequency traces, thus the adaptive time-frequency analysis is realized with an affordable computaional burden. Furthermore, an adaptive LPFT based multiple signal classification (MUSIC) algorithm is proposed. Compared with other time-frequency MUSIC algorithms, the proposed algorithm is more adaptable to signal patterns and has considerable advantages in the aspects of DOA estiamtion pricesion and multiple sources separation.