State space reconstruction is the basis of nonlinear analysis and the C-C method derived from the correlation integral is an efficient way to estimate the two parameters for state space reconstruction, the delay time τd and delay time window τw. For chaotic systems are sensitive to initial conditions, and measured data with finite number are noise-corrupted, the estimates of τd and τw with C-C method are fluctuant. In order to reduce estimate deviations, similar to the average method in spectral estimation, a time series was divided into several segments and each one was used to estimate corresponding values and the averages were taken as the estimates of τd and τw.This method differs from the ones in literature which used the entire series for estimate. The estimates with noise corrupted data and the effects of series length on the estimates were discussed. Numerical simulation showed that method is effective and reliable for estimates of τd and τw.
