Sampling and Quantization - Digital Speech: Coding for Low Bit Rate Communication Systems - page 25

3.3 Scalar Quantization

Quantization converts a continuous-amplitude signal (usually 16 bit, rep-
resented by the digitization process) to a discrete-amplitude signal that is
different from the continuous-amplitude signal by the quantization error
or noise. When each of a set of discrete values is quantized separately the
process is known as scalar quantization. The input-output characteristics of
a uniform scalar quantizer are shown in Figure 3.2.
Each sampled value of the input analogue signal, which has an infinite
range (16 bit digitized), is compared against a finite set of amplitude values
and the closest value from the finite set is chosen to represent the amplitude.
The distance between the finite set of amplitude levels is called the quantizer
step size and is usually represented by . Each discrete amplitude level x i
is represented by a codeword c(n) for transmission purposes. The codeword
c(n) indicates to the de-quantizer, which is usually at the receiver, which
discrete amplitude is to be used.
Assuming all of the discrete amplitude values in the quantizer are repre-
sented by the same number of bits B and the sampling frequency is f s ,the

Output

111

y8

110

y7

101

y6

100

y5

x4 x7

x5 x6

x1 x2 x3

011

y4

Input

010

y3

001

y2

000

y1

Figure 3.2 The input-output characteristics of a uniform quantizer