Given a zero mean stationary Gaussian process{xₙ}, it is shown that the autocorrelation can be estimated byR̂ₓₓ(j) = CN ∑min{i=1}max{N} xᵢ sign(xᵢ₊ⱼ)whereCN={π}{2N²} ∑min{i=1}max{N}|xᵢ|. This method is attractive sinceCNneeds to he Computed only once, and∑min{i=1}max{N} xᵢ sign (xᵢ₊ⱼ)can be computed by additions only. Moreover, the shape of the autocorrelation given byR̂ₓₓ(j)/CNcan be computed by additions only.
No takes yet. Share an insight, caveat, or question.
D. Hertz (1982) studied this question.