On the feasibility of voice input to an on-line computer processing system

  • Authors:
  • Howard A. Elder

  • Affiliations:
  • Cornell Univ., Ithaca, NY

  • Venue:
  • Communications of the ACM
  • Year:
  • 1970

Quantified Score

Hi-index 48.22

Visualization

Abstract

An on-line digital computer processing system is considered in which an ordinary telephone is the complete terminal device, input to the computer being provided as a sequence of spoken words, and output to the user being audio responses from the machine. The feasibility of implementing such a system with a FORTRAN-like algebraic compiler as the object processor is considered. Details of a specific word recognition program are given. This technique depends on three simplifying restrictions, namely, a “small” vocabulary set, “known” speakers, and a “moment of silence” between each input word. Experimental results are presented giving error rates for different experimental conditions as well as the machine resources required to accommodate several users at a time. The results show that at this time it is both economically and logically feasible to handle at least 40 users at a time with an IBM 360/65 computer.