An attempt of automatic recognition of some Russian words

A hierarchially organized computing recognition system is described. The primary dynamlc spectrogram is obtained by means of 23 1/3 - octave filters. The algorithms of the lowest recognising levels deal with finding out the phonetic features in the spectrogram. At the highest recognizing levels the nonunanimous string of phoneme codes is converted by the lexical algorithm into the sequence of words of the restricted vocabulary. The grammatical and semantical algorithms select those of them which are in accordance with the rules of the phrase construction. 150 Russian words pronounced isolatedly and in simple phrases have been recognized by the computing system with 90% effectiveness.