Hasan, A. K. M. Rashidul <br/><br/>
Feature extraction through spectral analysis for speech recognition 

 - Bangkok :  Asian Institute of Technology,  1995 
 - 56 leaves 
 - Research studies project report ;  no. TC-95-14 .
 - Asian Institute of Technology.  Research studies project report ;  no. TC-95-14 .
<br/><br/> A research study submitted in partial fulfillment of .the requirements for the deg1·ee of  Master of Engineering, School of Engineering and Technology 
<br/><br/>Research Studies Project Report (M. Eng.) - Asian Institute of Technology, 1995 
<br/><br/> This study presents a method for extracting speech features through spectral analysis for  speech recognition purposes. After a brief overview of all the major spectral analysis techniques  available, LPC-Derived Bilinearly Transformed Cepstral Coefficients have been proposed as the  working coefficients. A program has been developed in MATLAB environment for implementing  the proposed algorithm and using real Thai speech (vowel parts of Thai syllables 'THI' and  'KHA:'), two codevectors are calculated. Through some calculations (e.g., mean or centroid,  distance etc.), threshold values have been set for evaluating the performance of the generated  codevectors. 
<br/><br/><label>Subjects--Topical Terms: </label><br/>Spectral analysis (Phonetics)<br/>Speech perception