Method and apparatus for recognizing unknown spoken words by feature extraction and comparison with reference words

摘要

Prior Art: Speech recognition takes place in several successive phases. Each word is separated into elements of equal length (20 ms), "frames", and each element is represented by a vector consisting of 8 coefficients. After word-end recognition, the word is successively compared with all reference words.;Technical Problem: Very time-consuming, so that comparisons with only few reference words are possible (small vocabulary).;Basic Idea: Better utilization of hardware by use of common hardware for all phases and by avoidance of idle time.;Solution: Use of a Cellular Array Processor whose processing elements operate predominantly as filters in the feature extraction phase and each compare two vectors in the comparison phase. By gradual transitions between ends and beginnings of words, the capacities of all processors are equally utilized even there. The exchange of data is practically limited to an exchange between adjacent processors.