## A Method To Write A Perfect Character Sketch

The difference is a quantizer entrance, and designates its exit. The size of a quantized error of prediction is coded by sequence of binary symbols and transmitted through the channel in point of reception. The quantized mistake is also summarized with the predicted size to receive.

Thus, is the weighed linear combination of the M counting, and are prediction coefficients. Sizes get out so that to minimize some function of a mistake between and. Let's illustrate the aforesaid on RS piece:

For function evaluation it is necessary to determine summation limits by n: where N – number of counting in the RS segment, and M - number of the counting necessary for calculation of coefficients of a prediction (M + - counting. Means, the first predicted value will register so: where n = M +

where - unrationed short-term AKF. As definition of function is reduced to calculation of AKF, such method is called autocorrelated. When using this method we receive the displaced estimates of coefficients of a linear prediction (however, at M <

Instead of use of block processing for finding of coefficients of the predictor { } as it is described above, we can adapt the predictor's coefficients pootschetno, using algorithm of gradient type which we also will consider.

It is necessary to tell that such algorithm meets at very large number of iterations, generally, at the number of iterations striving for infinity. Therefore it is also necessary to be set before calculations by an admissible error which can suit us.

When using a kovariatsionny method not displaced estimates of coefficients of a linear prediction, that is E { ak } = ak.ist, where ak.ist – true values of coefficients of a linear prediction turn out.

measurement, coding and transfer on the reception party of the RS parameters in which already on the reception party synthesis of this (artificial) RS is made. Such systems are called vocoder (Source Codec);

As coefficients of reflection and coefficients of a prediction are calculated within the same procedure of algorithm of Levinson-Durbina, they can be expressed the friend through the friend. Let's give these algritma here.