One of the most important properties in gene expression is the stochasticity. Gene expression process is noisy and fluctuant. In this paper, the quantitative analysis of noisy time-series gene expression data on inference of gene regulatory networks is performed. We propose a two-step algorithm to solve the problem. In the first step, B-Spline is introduced to interpolate between data points. In the second step, Kalman filter or H∞ filter is introduced to infer the gene structure. If the statistical noise is known, Kalman filter is applied; Otherwise H∞ filter is applied. Both synthetic data and real experiment data are used to evaluate the procedure.