著者
新美 康永 小林 豊 浅見 俊幸 三木 豊
出版者
一般社団法人情報処理学会
雑誌
情報処理 (ISSN:04478053)
巻号頁・発行日
vol.18, no.5, pp.p453-459, 1977-05-15
被引用文献数
2

This paper describes a highly predictive speech recognition system, developed as a voiceinput ptogramming system, in which a modified version of 'BASIC', named 'SPOKEN BASIC 1', is used. The system consists of four major components; acoustic processor, lexical matching procedure, syntactic processor and semantic processor. The acoustic processor transforms incoming speech signals into a sequence of labeled segments. The syntactic processor makes grammatical predictions using a left-to-right parsing scheme and a depth-first or a breadth-first tree search. The semantic processor refines those grammatical predictions and sends predicted words to the lexical matching procedure, which correlates them with the sequence of labeled segments. 142 sentences uttered by four male speakers were processed through the present system. It responded as follows: 116 sentences (81.7%) were correctly recognized, 19 (13.4%) incorrectly recognized and the others rejected.