提出一种对缓慢变化的单样本非平稳状态的随机过程各个时刻的方差进行估计的新方法。它利用三次样条插值的原理，先在两相邻数据间进行一定数量的插值，并估计出插值给数据带来的附加误差，然后将这一误差从计算结果中修正掉。实验结果表明，使用这一方法估计的方差与实际符合状况良好。 A new method is proposed to estimate the variance of each stochastic process with slowly changing single sample nonstationary states. It uses the principle of cubic spline interpolation, first a certain number of interpolation between two adjacent data, and estimate the interpolation to the data to bring the additional error, and then the error corrected from the calculation results. Experimental results show that the variance estimated using this method is in good agreement with the actual situation.