Unformatted text preview: ECES-434/558 Assignment Four Optional Due Friday December 4, 2009 at 1 PM 1. Recognize speech or Wreck a nice beach! Starting with the lpcFrames.m function, examine the 5 speech sound files provided on the course website: female1.wav male1.wav female2.wav male2.wav female3.wav The files contain different people saying the phrase “Critical equipment needs proper maintenance”. They are sampled at 16 kHz. (a) Early attempts at developing systems for automatic speech recognition (ASR) attempted to utilize formant frequencies to identify different phonemes. We can estimate the formant frequencies from the pole locations of the LPC polynomial A ( z ) (the columns of the output from lpcFrames ), using the functions roots and angle . Unfortunately, roots will only work on one column at a time, so write a MATLAB script to convert each column into it’s complex poles, and then find the angle of each pole. Of course, angle produces an angle in radians (where π is the Nyquist frequency), so also convert those to Hz and plot the pole angles vs. frame number on a single graph (perso also convert those to Hz and plot the pole angles vs....
View Full Document
- Spring '09
- Hertz, Speech recognition, complex conjugate, formant frequencies