While identifying the parameters of IMFs from Empirical Mode Decomposition, by Hilbert-Huang Transform, a piece of approximately linear data segment is necessary for a specific result. The select of the data segment will directly influence accuracy of the parameters. The time for getting the approximately linear data segment is required to be as short as possible. The paper uses Least Square Series-piecewise Linear Fitting method to divide data into pieces, then chooses several pieces with the highest goodness-of-fit, and takes each median as basis to change the length, for higher goodness-of-fit. The needed data segment is achieved in the case that this data segment can still reflect the inherent parameters. This paper brings some examples to verify that the approach is feasible and exact.