NCJ Number
118116
Date Published
1987
Length
5 pages
Annotation
Automatic speaker verification is a process by which a machine authenticates the claimed identity of a person from his voice characteristics.
Abstract
The Rensselaer Polytechnic Institute has researched this technology with the goal of developing a fast, simple, reliable, and inexpensive speaker verification system based around the Texas Instruments TMS32020 fast signal processing microprocessor. Several aspects of the research problem have been pursued including the feature space, feature selection, and classifier design. Better recognition accuracy can be obtained by combining the various families of features to form a larger feature space. To implement the speaker verification algorithm on real-time hardware, a new feature selection criterion has been developed. Details of this feature are discussed along with experimental results on a population of ninety speakers. 4 figures, 10 references. (Author abstract modified)