This is, possible because the data correlated,

and therefore this predictability, and more specifically the

linear predictability we'll be discussing here, has to

do something with the autocorrelation of the signal.

Something that we covered earlier in, the

course for example when we covered window filtering.

[BLANK_AUDIO].

The basic principle of DPCM or predictive coding is that,

anything that can be predicted from past values, can be

done so both at the, encoder and decoder, and therefore

what we need to send to the decoder is the, unpredictable

part of the signal, which is just the prediction error.

[SOUND] As we will see,.

The prediction coefficients, denoted by a, will be the result of the solution

of this set of, equations, so called normal equations, that, relate

autocorrelation values again to the, the near prediction coefficients.

So, let's see the structure of the predictor and, encoder decoder.

And also, derive these so-called normal equations.

[SOUND].

We show here the broad diagram of a DPCM, Encoder.

X of n is the signal to be, encoded is shown here as a one-dimensional signal.

X cut of n is the predictive, value of the

signal at the same, time instance or the same spatial occasion.