Date of Award
Spring 5-2017
Document Type
Thesis
Degree Name
Master of Science (MS)
Department
Mathematics
First Advisor
John C. Chrispell, Ph.D.
Second Advisor
Frederick A. Adkins, Ph.D.
Third Advisor
Yu-Jo Kuo, Ph.D.
Abstract
Each individual's voice has unique characteristics that can be distinguished from someone else. When two people vocalize the same pattern of speech, we can easily identify that there are two different speakers. To establish a working framework in converting a voice, a sample phrase of a desired voice will be compared with the same sample phrase of the source voice. Constructing a basis for each sample will allow for frequency analysis between the two recordings. From this comparison, a mapping will be identified from the given input to the the desired output. Using two individual voices: whenever the desired phrase is recorded, the data undergoes a transformation as key frequencies are mapped to the desired output and results in the phrase spoken with the desired sound. When determining the key characteristics of each audio signal fast Fourier transforms will be used. Once a mapping is identified, future input signals will be mapped to the desired output frequency using both a FFT and an IFFT (inverse fast Fourier transform) to produce the desired modulation of the given input.
Recommended Citation
Lilla, Nathan, "Mathematical Methods for Voice Transformation" (2017). Theses and Dissertations (All). 1485.
https://knowledge.library.iup.edu/etd/1485