Investigation of speaker identification based on nasal phonation.

• Main Author: Young, Robert Bryant.
• Other Authors: Campbell, J.D.
• Language: en_US
• Published: Monterey, California. Naval Postgraduate School 2012
• Online Access: http://hdl.handle.net/10945/15783

This thesis investigates the possibility of Speaker Identification through the use of Nasal Phonation. Short segments of a restricted set of words from one speaker were sampled, processed, and the resulting vector is used to represent the speaker. Representative vectors were formed for several speakers and correlated with vectors representing individual words from "test" speakers. The magnitude of the correlations of the word vectors with various speaker vectors were used to identify the speaker. This work expands on earlier work done in this field to the extent that it attempts to remove the subjective preparation of data and replace this instead with an objective process of computer mechanization. Some limited success was achieved and, just as important, critical problem areas are noted which, if improved upon as recommended, promise an improved identification capability. Two different word lists fundamental to the identification process were also investigated. Some data was obtained but it was not sufficient to suggest that one word list would be more productive than the other when used as the basis for speaker identification. Recommendation is made to pursue further research in Speaker Identification using computer programming established during work on this thesis.