Accession Number : ADA194153

Title :   An Overview of Technology for Spoken Interaction with Machines (Une Introduction a la Communication Vocale avec les Machines).

Descriptive Note : Aeronautical note,

Corporate Author : NATIONAL AERONAUTICAL ESTABLISHMENT OTTAWA (ONTARIO)

Personal Author(s) : Hunt, M J

PDF Url : ADA194153

Report Date : Feb 1988

Pagination or Media Count : 38

Abstract : This report provides a non-mathematical introduction to speech input and output technology. It is divided into three parts. The first presents necessary background information on speech: on its nature, its production and perception, and on methods of analysis and coding used in speech I/O. A central message is that our subjective impression of speech is misleading and causes us to underestimate the complexity of speech communication. The second part is concerned with speech output and discusses the trade-offs that must be made between the quality and flexibility of the speech generated and the complexity and storage requirements of the speech output system. The final - and longest - part of the report deals with speech recognition. Arguments are presented in favor of statistical rather than rule-based approaches to speech recognition. The categories of recognizer currently available and the algorithms they use are briefly described, with the general conclusion that the performance obtained depends critically on the training process: on the type and quantity of the training material and on the amount of information derived from it. Three more detailed sections cover spectral representations and distance measures, the particular set of representations classed as auditory models, and techniques for handling noise and distortions. The last section discusses the difficulties of specifying recognizer performance, and recommends that all performance measurements should be treated with circumspection.

Descriptors :   *MESSAGE PROCESSING, *SPEECH RECOGNITION, *VOICE COMMUNICATIONS, ALGORITHMS, CODING, HANDLING, HEARING, INPUT, INSTRUCTIONAL MATERIALS, INTERACTIONS, MODELS, NOISE, OUTPUT, PERCEPTION, PRODUCTION, RANGE(DISTANCE), RECOGNITION, REQUIREMENTS, SPECTRA, SPEECH, STORAGE, TRAINING, NATO

Subject Categories : Voice Communications

Distribution Statement : APPROVED FOR PUBLIC RELEASE