LINEAR PKIODU'TION IN TELEVISION 



7G7 



COMPUTER 



s,(t) I 



— *^ — I- 



I " 



S,(t)-Sp(t) 



.- fe_ 



Sp(t)l 



PREDICTOR 



Sp(-t) 



s,(t) 



Fig. 2 — Decorrelator and correlator showing reversible nature of this method 

 of removing redundancy. 



bases its prediction on the "previous frame," and we are transmitting 

 a "still," there will be no surprises after the first frame and consequently 

 no output signal. Certainly it is redundant to send the same picture 

 more than once. 



linear prediction pro\ades an easily instrumented means of removing 

 redundancy. With linear prediction the next signal sample is simply 

 the sum of the previous signal samples, each multiplied by an appropriate 

 weighting factor. The best values for these weighting coefficients de- 

 pend on the statistics of the signal. 



Fig. 3 is a block diagram of a decorrelator employing linear predic- 

 tion. The delayed versions of the input signal can be obtained from 

 taps along the delay line. The weighting coefficients for each of the 

 delayed signals are selected by loss in their respective paths as shown 

 by the amplitude controls. The polarity of each signal can be determined 

 by the switches. The output is simply the sum of these weighted signals. 



If we consider the signal on a continuous basis (not quantized or 

 sampled), linear predictors can be characterized as ordinary linear 

 filters used to predistort the frequency spectrum of the signal. As such, 

 they can be designed in the frequency or time domain. However as will 



Fig. 3 — General block diagram of decorrelator employing linear prediction. 

 Linear prediction bases its prediction on the weighted sum of previous signal 

 samples. 



