Parameters Estimation for Colored Non-Gaussian Background in Signal Detection
Feng Liu, Wang Pingbo, Tang Suofu, Cai Zhiming · 2010
LS-EM algorithm can estimate Gaussian mixture autoregressive model (GMAR) parameter, which is one of the most efficient models for fitting PDF/PSD of non-Gaussian colored processes, especially interference background of detections. But its operation amount is too huge to be applied in real time. A modified LS-EM algorithm (MLS-EM) is proposed, which aborts the unnecessary feedback and coupling link in order to enhance the estimating speed. is faster than LSEM despite of its efficiency is lower a little. Applied in CPWG, the asymptotically optimal test of weak signal in the presence of colored non-Gaussian interference background, MLS-EM can save almost half of calculating time while its detecting performance is very close to LS-EM.