Training Methodologies for Dependent Speech Recognition (SR) Systems
Abstract
An experiment was conducted to determine whether a dependent SR system would perform with different accuracies given different ways in which it was trained. The experiment used a SR system (Voice Navigator) which is based on Dragon Systems, Inc. (proprietary) technology. Fifteen subjects trained three different voice patterns each and conducted four tests to compile statistics about the recognition accuracy for each pattern. The experiment was successful and demonstrated that the training method used can have significant impact on the performance of a dependent SR system. This thesis discusses the research methodology, reviews and analyzes the data collected, and states conclusions drawn about the particular dependent SR system used in the experiment.
Document Details
- Document Type
- Technical Report
- Publication Date
- Mar 01, 1991
- Accession Number
- ADA242983
Entities
People
- Richard L. Miller
Organizations
- Naval Postgraduate School