摘要
提出了一种估计白噪声中正弦信号个数的新方法—比率准则（ＲａｔｉｏＣｒｉｔｅｒｉｏｎ），简称ＲＣ准则。仿真结果表明：这一方法在用于估计白噪声背景下正弦信号个数时，能够给出正确结果，避免了信息类准则估计信号个数偏高的缺陷，明显地优于信息类准则。
Inaccurate estimate of number of sinusoidal sisnais in white noise has proved to be a problem difficult to solve. The 1995 'RELAX' method of Li et alL5], though dealing with better processing of stunals in white noise, did not seek to obtain the correct number of signals. We now present a method, called 'Ratio Criterion(RC)' by us, that can give a correct estimate. The key equations of our RC method are eqs. (5) through (7). These key equations are based on three information theoretic criteria well known in signal processing theory (FPE for Final Prediction Error, AIC for An information Criterion, and BIC for Bayes Information Criterion). Instead of minimiZing FPE (k), AIC (k ) and BIC (k ) as was previously done, we minimize RCFPE(k), RCAIC (k ) and RCBIC(k ) to obtain the estimated number of sinusoidal signals. For white noise with one to five signals we perform simulation calculations and all obtain satisfactory results even for SNR as low as 10dB. Fig. 3, for three signals show that for 100 samples (N=100), minimizations of the three ratios in eqs. (5) through (7) ail sive correct estimated number of signals. Fig. 4,for three signals,show that with SNR decreasing below 10dB,the probability of correct estimate of number of signals decreases.
出处
《西北工业大学学报》
EI
CAS
CSCD
北大核心
1996年第2期304-308,共5页
Journal of Northwestern Polytechnical University
关键词
信息准则
比率准则
白噪声
正弦信号
information theoretic criteria, ratio criterion, estimate, number of sinusoidal signals, white noise
