SPOKEN WORD RECOGNITION SYSTEM FOR UNLIMITED SPEAKERS USING GROSS PATTERN OF SPECTRUM AND LINGUISTIC INFORMATION.

A description is given of the newly developed spoken word recognition system which is composed of the following four stages. In the first stage, seven parameters are extracted every 10 msec from the outputs of the filter bank. The parameters are the frequencies (channel numbers) of three spectral local peaks, the speech power and three parameters expressing the gross pattern of the spectrum. In the second stage, the segmentation and phoneme recognition are carried out. In the third stage, the errors in the segmentation and phoneme recognition are corrected by means of phoneme connecting rules. And, in the last stage, the item of the dictionary having maximum similarity to the recognized phonemic sequence is chosen. Every item of the dictionary is written in phonemic symbols which is transformed from the word written in Japanese ″Kana″ letters by simple rules.