Experimental text‐to‐speech system for the handicapped

Abstract

Systems for generating speech from unrestricted text have recently been brought to a level appropriate to several applications. Since some applications, e.g., in the rehabilitation area, are radically new, user system requirements remain to be specified. To that end, a stand‐alone text‐to‐speech system has been implemented with enough flexibility as to programming and interface to allow high speech quality and easy adaptation to different needs. The speech is generated from text by a set of explicit rules [Carlson and Granstrom, IEEE ICASSP (1976)]. There is a branching capability, so that different sets of rules are applied to different items, e.g., numbers. The rules are augmented by a user programmable lexicon in which the user could introduce codes for word sequences. Several system applications will be discussed: as a general text‐to‐speech converter in the telephone network, in automatic recording of talking books, as a vocal output for computerized information systems. In particular, experience in the use of the system as a personal speech prosthesis for a nonvocal CP patient will be reported.