Research on Voice Authentication
Abstract
The purpose of this study was to develop voice authentication techniques which can be used over packet switched networks without the limitations of a single fixed reference phase. The results of this study were significant in several respects. First of all, the largest data base of controlled conditions but extemporaneous human speech in existence was developed for the project. Secondly, a real-time processing capability was developed for processing this data base of 40 hours of continuous speech. Finally, with the restriction of clean speech (not processed over a telephone or with noise), text-independent speaker recognition results nearly matching those for text- dependent studies were achieved by averaging over 30-40 second speech segments.
Document Details
- Document Type
- Technical Report
- Publication Date
- Feb 01, 1978
- Accession Number
- ADA104316
Entities
People
- John D. Markel