Hi,
answer to query 2:

Recall the LP formulation
s^(n) = sum(ak. s(n-k)).
The error signal is
e(n) = s(n) - s^(n)

Or,
E(z) = S(z) A(z), where A(z) is the all-zero or FIR filter or inverse filter.
The above equation is LP analysis equation, where we derive the
error signal form the speech signal.

The synthesis equation is
S(z) = E(z) H(z), where H(z) = 1/A(z), is an all-pole filter, also
called as the synthesis filter or LP filter .
Ans. to query 1:
The idea of preemphasis during LP analysis is to reduce the spectral dynamic
range. In particular, the lower frequencies have significant energy, and dictate the
shape of spectral envelope as estimated by LP analysis. To reduce this effect,
use x(n) = s(n) - a.s(n-1). Use a close to 1, say 0.99.
This is equivalent to a high pass filtering, though not with a sharp cutoff.
This somewhat deemphasizes the low frequency components and ensures that
the estimate of the spectral envelope, as obtained by LP analysis, is not unduly
influenced by the large spectral magnitude in low-frequency range.

-Guru

--- On Thu, 8/7/08, phitenk wrote:
From: phitenk
Subject: [speechcoding] preemphasis filter, FIR, IIR
To: s...
Date: Thursday, August 7, 2008, 6:55 PM

hy all,

1. I want to ask, how to find the preemp coefficient 15/16=0.98.. ??

2. why in LPC anlysis we using FIR & in LPC synthesis we use IIR

filter? what is the reason?

thanks,,