How about first supress noise then compute LP?
--- zhaohanwu <zhaohanwu@zhao...>ะด
> I'm looking for a robust LP method in
noisy
> environment (around 0dB),
> which method can get better LP coefficient in harsh
> noisy background.
>
> here is one method named "WLSP" proposed by Alku and
> Backstrom "All-
> pole modeling technique based on the Weighted sum of
> the LSP
> polynomials" in Proc. IEEE ICASSP'02 orlando, 2002.
> Have anyone test it?
>
> i want use this method the extract robust LP coeff,
> so as to improve
> the robustness of MELP algorithm. how about this
> idea?
>