Abstract: From the view of trick recognition，features of signal decomposition structure in existing frequency estimation methods are analyzed，based on which，frequency-division-domain decomposition structure is proposed.It organizes several simple signals corresponding to different time into one compound signal.So the given information about time domain can be made use and it will form information accumulation effect to suppress interference frequency and weaken impact noise effectively without incensement in calculation.To match usage of frequency-division-domain decomposition structure and improve real-time performance of analysis，frequency-division equilong fusion algorithm suitable for equilong signals and indirective iterative computation is adopted，and mathematical justification of algorithm correctness is provided.Simulated experiments are carried out under many kinds of application environment，in which better noise immunity，real-time performance and higher frequency estimation precision are shown.

Key words: frequency estimation, frequency-division-domain, DFT, information fusion, iterative computation

摘要： 从特征识别角度分析了现有频率估计方法的信号分解结构的特点，并提出了一种分频域分解结构，将对应不同时间的多个单频信号融合构成一个组合信号，以利用时域已知信息并形成信息积累作用，能有效抑制干扰频率和削弱冲击噪声且计算量增加较少。为配合分频域分解结构的使用和提高分析过程的实时性，采用了分频等长融合算法和间接迭代计算，并进行了方法正确性的数学证明。针对多种应用环境状态进行了仿真实验，结果表明该方法抗噪性和实时性较好，频率估计精度比现有方法有较大提高。

关键词: 频率估计, 分频域, DFT, 信息融合, 迭代计算