Algorithms for Communications Systems and their Applications, 2nd Edition

Chapter 5Vector quantization

Vector quantization (VQ) is introduced as a natural extension of the scalar quantization (SQ) [1] concept. However, using multidimensional signals opens the way to many techniques and applications that are not found in the scalar case [2, 3].

5.1 Basic concept

The basic concept is that of associating with an input vector , generic sample of a vector random process , a reproduction vector chosen from a finite set of elements (code vectors), , called codebook, so that a given distortion measure is minimized.

Figure 5.1 exemplifies the encoder and decoder functions of a VQ scheme. The encoder computes the distortion associated with the representation of the input vector by each reproduction vector of and decides for the vector of the codebook that minimizes ...