PhD

dissertationClinical computer applications are becoming increasingly important in health care delivery. The expense of video terminals limits access to the power of the computer. Voice response, using a common telephone as a computer terminal, increases accessibility and is shown to be a workable means of communication in a hospital environment. Current methods of processing speech for voice response are reviewed. Analog, digital, and parametric methods are compared and evaluated for use in a hospital environment. Great importance is given to the need for economy, large vocabulary with low storage requirement, flexibility of application, and compatibility with the existing computer system at LDS Hospital. Reasons for choosing Infinitely-Clipped speech are presented. Investigations into methods of enhancing the intelligibility of Infinitely-Clipped Speech are reviewed. Peripheral hardware, constructed to code the reproduce clipped speech in real time outside the computer, is described in detail. A coding scheme, which preserves intelligibility while reducing data storage requirements, is presented. VOTRAX, a commercial speech synthesizer, is compared to the Infinitely-Clipped Speech system. Software systems to store and retrieved coded equivalents of spoken words are described. Distinct dictionary constructions are presented to meet the requirements of the different synthesizers. Provisions were made to allow users to control repetitions of data input requests and output messages, in addition to error correction and input data verification procedures. Software and hardware were designed to be compatible with existing video display terminal communication protocol. Blood Gas Test Requests, Blood Gas Analysis Reports and Patient Problem List Review applications were adapted from existing video terminal implementations for voice response trials. The selection of these applications was based on the large vocabulary required, the limited input and output message length use, and the need for infrequent use from diverse locations within or outside the hospital. Communication of Blood Gas Requests and Reports expanded from one to eight nursing divisions. Intelligibility tests were performed with participants from the eight nursing divisions in their actual working environments. Results for both ICS and VOTRAX processing approached 90% for isolated words taken from the Blood Gas applications. The importance of familiarization with the peculiar accents of the systems is noted. Use and acceptance of voice response in the chosen applications is discussed. Future applications are suggested