摘要
基于ＨＭＭ的多频率线跟踪算法，能在很低的ＳＮＲ环境下工作，但存在一个量化误差，此误差的量级高于所给测量条件下的 Ｃ－Ｒ下界所对应的误差量级（Ｎ－３／ ２）．本文提出，在 ＨＭＭ跟踪器输出的基础上，进一步实施极大似然估计以提高频率估计精度．为了极大地减轻系统的存贮数据负担，文中提出只选取测量信号ＦＦＴ的很少一部分，共Ｌ个数作为依据来实施估计．给出了估计算式和Ｌ的计算方法和仿真结果．
HMM based multiple frequency line tracker works well under very low SNR, except for a quantization error, the mean squares value of which may be much larger than the CramerRao lower bound. To achieve an accuracy near this limit, a fine frequency estimator is presented on the basis of the HMM tracker output, which has already given the interval within which the true frequency exists. To considerably reduce the system's data accumulation, this maximum likelihood estimator is performed based on only a very small number(L) parts of the FFT of the measurement data, but still gives good results. The computation of L is presented. Simulations support all these results.
出处
《上海大学学报（自然科学版）》
CAS
CSCD
2000年第5期377-383,共7页
Journal of Shanghai University:Natural Science Edition
基金
国家自然科学基金资助课题!(69672016)
关键词
频率线跟踪
隐马尔可夫模型
频率估计
信息处理
frequency line tracking
hidden Markov models
SNR threshold
Cramer--Rao lower bound
