SignalUtils::ArYuleWalker Function

Yule-Walker (autocorrelation) autoregressive parameter estimation.

Estimates the coefficients of an autoregressive model of Src by solving the Yule-Walker normal equations on the biased autocorrelation via the Levinson-Durbin recursion. The model assumes x[n] = -sum_(i=1)^p a_i x[n-i] + e[n] and the returned prediction-error filter is A(z) = 1 + a_1 z^(-1) + ... + a_p z^(-p), so A has the form [ 1, a_1, ..., a_p ].

CONVENTION: the Order argument is the number of autocorrelation lags, which equals the length of A. For an order-p model (filter length p+1) pass Order $= p+1$. This differs from ArBurg, ArCovariance and ArMCovariance, whose Order argument is the model order p and which size A to p+1 internally. Order is clamped to Src.Length. The roots of $A(z)$ are the estimated poles; a stable model has all roots inside the unit circle.