Analysis by Synthesis LPC Coding - Digital Speech: Coding for Low Bit Rate Communication Systems - page 196

Observed
Signal

+

s(n)

^

s(n) −

System
Model

+

Error
Signal
e(n) Improve Model

Parameters

Figure 7.1 General block diagram of analysis-by-synthesis closed-loop analysis

and represented in some form, e.g. the time or frequency domain. Then a
theoretical form of the signal production model is assumed, as depicted in
Figure 7.1. The model has a number of parameters which can be varied to
produce different variations of the observable signal. In order to derive a
representation of the model that is of the same form as the true signal model,
a trial and error procedure can be applied. By varying the parameters of
the model in a systematic way, it is possible to find a set of parameters that
can produce a synthetic signal which matches the real signal with minimum
possible error (assuming the model is valid to begin with). Therefore, when
such a match is calculated, the parameters of the model are assumed to be the
parameters of the true signal.
The AbS procedure outlined above was applied to speech processing in the
earlier days of formant estimation [7] but, because of its obvious complexity, it
was not re-applied until Atal outlined the basis of Multi-pulse LPC (MPLPC)
in [8] for low bit-rate coding. In Atal's work, the time-domain representation
of speech was used and a model very similar to the conventional source-filter
model was selected. However, AbS with other domains and models are
equally applicable [9]. In the following sections a unified presentation of the
various AbS-LPC schemes using Atal's modelling is described.

7.2 Generalized AbS Coding

The basic structure of an AbS-LPC coding system is illustrated in Figure 7.2.
There are basically three blocks in the model that can be varied to match our
true model and, hence, obtain a good synthesized speech signal: time-varying
filter , excitation signal and perceptually-based error minimization procedure .Asour