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． The HMM-based multi-frequency line-tracking algorithm can work in a very low SNR environment but has a quantization error that is orders of magnitude higher than the error magnitude corresponding to the C-R lower bound under the given measurement conditions (N -3 / 2). In this paper, based on the output of HMM tracker, maximum likelihood estimation is further implemented to improve the accuracy of frequency estimation. In order to greatly reduce the data storage system burden, this paper proposes only a small part of the measured signal FFT, the total number of L as a basis to implement the estimate. The calculation methods and simulation results of the estimation formula and L are given.