Ethnicity
Methods for Speech Recognition
A Project for the Spectral Analysis course ELEC 532.
Rice University, May 1998.
Group members
Jalal Benharouga
Ricardo von Borries
Daniel Ljunggren
Pamela Loughmiller
Mark Wagstaff
-
-
-
-
-
Morocco
Brazil
Sweden
Texas
U.S.A.
Table of Content
Introduction
Speech recognition and some applications
Use of different models
Power method
A "simple" method using windows
Based on the characteristic power-distribution
Formant Trajectory
What is formant?
Analyzing frequencies
Identify patterns
Cepstrum method
More advanced method, although easily implemented
Based on LPC analysis
Cepstral coefficients gives model
Result
Conclusions
Future work
Bibliography
Ethnicity group, May 4 1998