Abstract: In this paper, a novel numerical solution method, moving least squares (MLS), is employed to solve the problem of nonlinear reconstruction of multi-functional sensor with a view that least squares (LS) are restricted in global regression. Through studying the construction method and characters of interpolated function, basis function and weight function are selected reasonably to obtain the coefficients in trial function, and the reconstructed value of input signals is acquired. This paper presents an analysis of the effects of the parameters in MLS, such as the dimensions of basis function, the number of points in the support domain, and the coefficient of weight function. Comparisons are made between LS and MLS reconstruction data, whose relative errors are smaller than 15.3 % and 1.03 %, respectively. The results demonstrate that MLS is suitable for nonlinear regression of curves. Additionally, more points in the support domain or higher dimensions of basis function will greatly increase the reconstructed accuracy.