The authors have developed a speaker independent word recognition system for the purpose of giving extension number guidance to our laboratories. The system consists of a workstation, AD converter, PBX interface and speech synthesizer. The vocabulary consists of the names of persons and laboratories, and its size is about 300. The recognition process is based on three methods: continuous DP matching, multi-templates and SPLIT methods. The system announces the results of recognition and guidance of the extension number to the user by using a Klatt type speech synthesizer.