The Burg algorithm [1, 2] is a technique for fitting an AR model to a signal that is represented by a sequence of *N* measured samples, *x*[0] through *x*[*N* –1]. What sets the Burg method apart from other techniques for estimating AR model parameters are the assumptions made about signal values *x*[*n*] for *n* > 0 and for *n* ≥ *N*. The autocorrelation method described in Note 70 assumes that unknown values of *x*[*n*] are zero. The covariance method described in Note 71 makes no assumptions about unknown values for *x*[*n*], but uses an optimization strategy that is structured to use only the *N* measured values. FFT-based methods assume the same periodic extension of values that is implicit ...

Start Free Trial

No credit card required