Man-machine communications are playing an increasingly important role in today's technology-oriented society. The ability to automatically recognize words from a limited vocabulary set has particular significance for computer-based learning systems, where student input is generally limited to typed responses. Research and development work aimed at sutyding the acoustical properties of speech to determine which parameters offer the most reliable discrimination between words of a limited vocabulary, requires the careful design and construction of voice data acquisition and processing equipment. A portable recording and digitizing apparatus is described in detail, which permits selectable automatic segmentation of the natural speech signal prior to digitizing and processing at a central computer facility. The flexibility and accuracy of the data acquisition unit are illustrated by describing the complete recording and analysis procedure for a sample word.
This paper costs $33 for non-members and is free for AES members and E-Library subscribers.