Conventional Design #1

• Open-loop approach (OL)

• Design:

Input frames: X : {x0, x1, x2, …, xN}

Training Set: T : {t0, t1, t2, …, tN}

where tn = xn - P[xn-1], n = 1,2, …, N,

P is the predictor operator

• Optimize VQ for the training set

Previous slide

Next slide

Back to first slide

View graphic version

The simplest approach to the design of a predictive quantizer is what is called the open-loop approach (OL). In this case you base your prediction on original samples and not on decoded samples. This way, you can easily generate a training set of residuals and then design a quantizer for them.