Text independent speaker Identification system using fused mel Frequency feature sets based on Gaussian mixture model
Loading...
Date
item.page.authors
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
Speaker identification is the process of identifying the individual
newlineusing the individual s speech signal for applications like secure information
newlineaccess system voice based attendance entry system, forensic system etc
newlineConventional speaker identification systems use long duration speech signals
newlinefor identifying the speaker The difficulty with speaker identification using
newlineshort duration speech signal is in acquiring sufficient speaker specific
newlineinformation from the speech signal This thesis investigates five methods of
newlinespeaker identification by utilizing short utterance of speech signals Block
newlinetruncation on wavelet sub band processed speech signal cross lingual speech
newlineprocessing multilingual speech processing combined vocal tract and auditory
newlinesystems quadrilateral structure filter bank analysis are used to obtain efficient
newlineclosed set text independent speaker identification
newlineA narrow band noise robust speaker identification system presented
newlinein this thesis is based on the idea of using sub band processing followed by
newlineblock truncation to build the model of a particular speaker When a speech
newlinesignal is partly degraded by a narrow band frequency selective noise a part
newlineof the speech spectrum may still be uncorrupted This uncorrupted spectrum
newlineprovides an opportunity for improvement in speaker identification
newlineperformance if the speech signal is processed in sub bands
newline
newline