In this chapter we consider first some of the requirements which different types of multimedia application can impose on speech coders. Then a brief overview of the existing speech coding standards will be given. In Chapter 4, we considered linear prediction techniques which are, in fact, used by almost all modern speech codecs. However, in addition to the high predictability, speech signals have specific features which are used by speech coders and allow us to achieve very high compression ratios. These properties are characterized by such parameters as pitch period, linear spectral parameters, etc. We will explain the meanings of these parameters and describe standard algorithms for finding them.

The most important attributes of speech coding are: bit-rate, delay, complexity, and quality.

The bit-rate is determined as the number of bits required in order to transmit or store one second of the speech signal. It is measured in kb/s. The range of bit-rates that have been standardized is from 2.4 kb/s for secure telephony up to 64 kb/s for network applications. Table 7.1 shows a list of standardized speech coders. It includes their bit-rates, delays measured in frame size and lookahead requirements, and their complexity in Million Instructions Per Second (MIPS). Several coders standardized by the International Telecommunication Union (ITU) are presented.