In this work, an isolated speech recognition system designed for a digital exchange is introduced. The hardware structure of the system is given and the functions of the each module are explained. Software steps and the algorithrns are given, the training phase of the system and to build up the library in the system memory are discussed.