随着实验模态分析方法的广泛应用和不断发展，在工程应用中，研究者希望能减少限制条件，增加分析的可靠性。本文发展一种识别系统模态参数的新方法，本方法无需知道系统的输入信息且不用建立系统模型，仅通过对测试记录的响应信号进行频谱分析即可识别系统的模态参数，依据系统中各响应点的信号，先采用离散反卷积方法从各响应点信号中分离出谐波参数作为初步的识别结果，在此基础之上，结合频域空间曲线拟合的方法修正识别结果中受频谱混叠影响而偏差较大的谐波参数，再联合这些已提取的谐波参数得到系统的模态参数。通过对实际结构测试数据的分析，证明本方法只用输出数据识别的模态参数与模态分析软件用传递函数识别的结果一致。

With the broadest application and development of modal analysis method， the demands such as less restrictive conditions and to increase the reliability of an analysis have been increased for more convenience in engineering applications. This paper presents a new modal analysis method without input data and any analytical models. Harmonic parameters of a point response can be extracted by spectrum analysis using the measurements obtained from structural testing. Preliminary results calculated by discrete deconvolution techniques are further modified by the frequency domain space curve fitting method. Modal parameters of system just come from a combination of these corrected results. The performed analysis and validation of real data indicate that the approach is feasible for engineering. Therefore， this paper provides the effective estimation scheme of modal parameters for experiment research on the system dynamic performance.