In classical regression analysis for signal processing,the noise is usually assumed to obey standard Gaussian dis⁃tribution,but in reality,many signals present non⁃stationary characteristics such as noise autocorrelation. The common general⁃ized difference method fixes the correlation coefficient between two consecutive samples when making difference processing of the noise autocorrelation. However,it’s well known that the related extent of the adjacent time points isn’t certain. In this pa⁃per,an vector difference algorithm with inorganic adaptivity is put forward by combining the vector triangle rule of addition and subtraction with the generalized difference method,in which correlation coefficients can be automatically adjusted according to the signal’s own regular pattern. Finally,this method is applied to noise autocorrelation examples,whose results show that the vector difference algorithm is far better than the generalized difference method in the aspect of inorganic adaptive ability,and can describe the change rule of signals better.%使用经典回归分析对信号处理时通常假定噪声服从高斯过程，然而现实中许多信号呈现噪声自相关等非平稳特性。常用的广义差分法对噪声自相关做差分处理时，固定了连续两个样本间的相关系数，但是现实中相邻两个时间点样本的相关程度往往不是确定的。将矢量三角形加减法法则与广义差分相结合，开创性地提出具有无机自适应性的矢量差分算法，其相关系数根据信号自身的规律自动调整。最后，将该方法应用于噪声自相关实例，结果表明矢量差分算法比广义差分法的无机自适应能力更强，能够更好地刻画信号的变化规律。
展开▼
