Document Type

Thesis - Open Access

Award Date


Degree Name

Master of Science (MS)

Department / School

Electrical Engineering


This literature and experimental research was conducted in order to accomplish several objectives. First of all, the establishment of a historical background of speech recognition efforts was desired. This background would reveal the various basic approaches taken in the past to recognize speech. In addition, the disadvantages and contributions of each system would be noted in order to recognize and analyze the problems inherent in recognizing speech. Secondly, an analysis of the major problems blocking automatic speech recognition was desired not only to investigate these problems, but also to evaluate the state of their solutions. Thirdly, an experiment was performed to reduce a specific part of one of those problems, the differences which exist in the voice pitch of speakers. Finally, a more intangible objective of this research was to provide a basis of experience and knowledge from which a speech recognition system could be designed and built. The following investigation into the automatic recognition of

speech by machine was undertaken so that the fundamental problem of recognizing normal speech could be better understood and more easily solved. The survey of automated speech recognition systems was made because various approaches taken in the past were considered valuable when attempting to improve the recognition of speech by machine. The major problems nested within this fundamental goal were analyzed because they were crucial to the understanding and emphasis of past speech recognition efforts. Finally, an experiment was performed so that actual experience could be obtained with speech, the quantity being recognized and analyzed. In addition, the experiment was conducted in order to help solve the problem of pitch differences among speakers' voices.

Library of Congress Subject Headings

Automated speech recognition



Number of Pages



South Dakota State University