5.1 INTRODUCTION

Speech signals occurring on telecommunications links are the product of two-way conversations. It is customary for one talker to pause while the other speaks; thus an active speech signal is present on a transmission channel for only a fraction of the time. In addition, even when only one talker is speaking, pauses occur between utterances and there are times when the circuit is simply idle. Therefore, it can be expected that, on the average, speech is present for considerably less than 50% of the time. Measurements show that speech is present on a telephone channel approximately 40% of the time [Brady, 5.1], averaged over a large number of trunks. This chapter discusses digital time-assigned speech interpolation (TASI) [Daymonett, 5.2; Hashimoto, 5.3; Tygohounis, 5.4; Poretti, 5.5] without and with overload channels, variable rate delta modulation used with speech interpolation, and speech predictive encoded communications (SPEC) [Campanella, 5.6; Sevilli, 5.7; Suyderhoud, 5.8] techniques, which exploit these activity properties to reduce the information rate needed to handle a multiplicity of speech-carrying telephone channels.

Digital TASI and SPEC have been implemented using the customary 8-bit-per-sample pulse code modulation (PCM) format for 64-kb/s digital telephone transmission with either -law companding for T-carrier systems used in North America and Japan [5.9] or A-law companding for CEPT-32 used in Europe and most other countries outside of North America and Japan [CCITT, 5.9]. With such systems, the number of channels carried on a transmission facility can be...