A one‐boarded, out‐performance speech recognition system for speaker‐independent isolated words has been developed. This board consists mainly of the 4‐bit, 1‐chip microcomputer and newly developed speech recognition LSI [Ohgaet al., IEEE Trans. Consumer Electron.CE‐28, 263–270 (1982)]. The board takes multiple templates and the KNN rules to cover various tokens of many persons. A new clustering method CLS (clustering with shared group) is based on the conception of the SNN method and theK‐means iteration procedure [Rabineret al., IEEE Trans. Acoust. Speech Signal Process.ASSP‐27, 336–349 (1979)]. The experimental results showed more than 95% recognition accuracy not only for ten Japanese city names but ten Chinese city names. CLS can be easily implemented on the 16‐bit microcomputer system and it takes less than 20 min to make seven templates with the microcomputer system.