Adaptive quantization with prediction correction at the receiver

The paper discusses techniques of adaptive nonuniform quantization used at the transmitter for improving the SNR of the quantizer, where quantization step sizes are predicted at the receiver, thus increasing the channel capacity. Three different schemes for step size correction at the receiver are described: (1) forward estimation of the step size by the evaluation of the absolute ratio of adjacent samples, (2) backward estimation and transmission of the step sizes, and (3) SNR improvement through computer processing. A digital simulation, performed with first-order Gauss-Markov sequences (symbolizing speech and picture signals) as the input source, shows that the proposed schemes yield better SNR than most other quantization schemes.