Hardware and software of a voice recognition system are constructed. It's hardware consists of a voice recognition board and a host computer FM-7. The former is constructed using the set of LSI's for processing words spoken by specific speakers, and a small number of standard IC's. The set of LSI's are MC4760,μPD7761D and μPD7762G made by NEC. According to the 12 commands reserved by the processing unit μPD7762G for voice recognition, the system software is programmed with the FBASIC. According to the number of registered standard patterns and/or the repetitive number of speeches per word or syllable, the exprimental results of identification accuracy are shown. When the number of standard patterns and the repetitive number are two respectively, the average accuracy of identification is above 94%, which seems to be a comparatively satisfactory result.