Vowel Recognition

1
Vowel Recognition

• Class EE 623
• By
• Asmita Akerkar
• Ashwin Philar

2
Topics

• Introduction
• Implementation
• Functions Developed
• Enhancements
• Review
• References

3
Introduction

• Speech
• Sound production
• Modulation
• Voice recognition
• Acoustic signal Words

4
Implementation

• Formants
• Vocal cords
• Vocal tract
• Resonant frequencies
• Phoneme
• Smallest part of a word

5
Analog voice
Sample based vector
Normalized
Determine Power Spectral Density
Extract Formants
Comparison and recognition
6
Matlab Functions

• Vowel Detector Function
• Reads a sound file and stores it into a vector
• Calls the block function
• Block function
• Divides the signal in time domain into ten equal
  blocks.
• Selects the block with the maximum power content.

7

• Normalize Function
• Normalizes the selected block and plots the
  signal.
• PSD Function
• Uses the Yule-Walker method of determining the
  Power Spectral Density.
• Returns frequencies and their power content.

8

• Yule-Walker Method
• The input
• Sample-based vector.
• The output
• Estimate of the signal's power spectral density.
• Calculated at
• Nfft (N 1024) equally spaced frequency points
  in the range 0,Fs, where Fs is the signal's
  sample frequency (8000 Hz).
• ORDER is the order of the autoregressive (AR)
  model used to produce the PSD.

9

• Auto Regressive Model
• The simplest model for the vocal tract,
  consisting of linked cylindrical tubes, produces
  an all-pole transfer function.
• The AR model has only poles.
• The order of the model is a function of the
  sampling frequency.

10

• Peak Function
• Determines the frequencies at which the peaks in
  the PSD occur.
• These are the formant frequencies
• The first three formants are selected
• Vowel Guess Function
• Determines the distance of the input point from
  all the other vowel points.
• Minimum distance criteria is used to determine
  the spoken vowel.

11
Power Spectral Density
Vowel E
12
Enhancements

• Detect vowels within a word
• Detect vowels spoken by different people
• Detect vowels in different languages

13
Review

• What is voice recognition?
• How is it implemented?
• Matlab Functions
• Enhancements

14
References

• http//www.owlnet.rice.edu/, 03/04/2001
• http//www.studorg.utcluj.ro/giurgiu/jewels/recom
  /sle3.html, 03/04/2001
• http//www.cnel.ufl.edu/hybrid/courses/EEL6825-fal
  l96/node8.html, 03/04/2001