摘要 |
A device for sensing a drunken state of a driver by analyzing his voice, comprises a parameter extractor (106) for producing a parameter corresponding to a linear prediction coefficient-cepstrum coefficient extracted from a digital signal obtained by converting the sound of the voice command, a storage means (108) for storing a reference parameter extracted from the voice of the driver when not drunk, a comparator (110) for comparing a current parameter extracted by the extractor (106) with the reference parameter so as to produce a difference signal representing the deviation, and a decision circuit (112) for determining the driver to be drunk or not according to whether the deviation is above a reference value or not. The parameter extractor (106) generates the linear prediction coefficient-cepstrum coefficient after producing a linear prediction coefficient by taking an auto-correlation on the digital voice signal. <IMAGE> |