To provide a voice recognition device for Chinese which can make a necessary memory small by making recognition precision higher and uses the phoneme similarity vector at the beginning or end.

The voice recognition device for Chinese includes a voice signal filter which receives a voice signal to generate a filtered analog signal, a digital- to-analog(A/D) converter which converts the voice signal into a digital voice signal, a computer which receives and processes the digital signal, a pitch frequency detection device which detects characteristics of the pitch frequency of the voice signal and then recognizes the tone of the voice signal, a voice signal preprocessor which detects the end point of a syllable of the voice signal and then determines the beginning and end of the syllable, and a training part which trains a syllable model based upon trained parameters of a be ginning part PSV model and an end part PSV model.