- ME Spectral Analysis (Burg):
-
given a segment of signal - finite duration, estimate the
spectral density
-
spectrum estimated via autocovariance functions
-
compute the possible few autocovar's using given data,
estimate the rest using ME
-
resulting spectrum is most consistent with the available
autocovariance data
-
solution very similar to LPC covariance method
- other applications:
-
statistical language modeling (estimating trigrams and bigrams)
-
natural language processing (Ratnaparkhi) -- POS tagging
-
discourse modeling (Jurafsky)