In the past decade, progress in semiconductor technology has made it possible to implement mathematical algorithms for the efficient coding of speech. This is achieved by removing part of the redundancy which is present in speech signals. Many different approaches have been studied, but the one which has emerged as the most practical to realize in digital form while producing good quality speech is linear prediction coding. This article gives a basic background in speech coding followed by a brief description of the approach that was selected in developing the Vocodec which is a linear prediction coder/decoder system operating at a transmission rate of 2400 bit s** minus **1. The general approach for this development was to simulate the system in software as accurately as possible and then to build a stand-alone, multi-microprocessor based hardware for real-time duplex operation. This equipment is now being used for transmission experiments.