Implementation of Linear Predictive Coding (LPC) of Speech

1
Implementation of Linear PredictiveCoding (LPC)
of Speech

• Term Project by
• Komel Rauf
• Saba Hameed
• Mahinn Zahoor

2
Speech Modeling Non-stationary
3
Speech Modeling LTI Model
4
Speech Modeling LTI Model
5
Speech Modeling Source (voiced)
6
Speech Modeling Source (Unvoiced)
7
Speech Modeling
8
Speech Modeling Transfer Function
9
Radiation
10
Overview
11
Voice Segmentation
12
Voice Segmentation - Filtering and Windowing
13

Voice Segmentation - Silence Detection

14
LPC - Motivation

• Speech Difference Equation for a pth order
  filter

Want to minimize the mean-squared prediction
error
15
LPC - Autocorrelation

• If we assume that s(n) is zero outside the
  interval
• We need to solve the following set of linear
  equations

16
LPC - Autocorrelation
17
LPC - Autocorrelation

• In matrix form the set of linear equations can be
  expressed as

18
LPC-Levinson-Durbin Algorithm

• By exploiting
• Toeplitz structure of the matrix
• Particular structure of the right-hand side of
  the linear system of equation
• We can use the efficient Levinson-Durbin
  recursive procedure to solve this particular
  system of equations.

19
LPC - Gain Coefficient

• It can be shown that the gain coefficient is
  given by

Where En is the minimum mean squared error
prediction and is given by E(p) from
Levinson-Durbins Algorithm.--We will transmit G2.
20
LPC Algorithm
21
Pitch Detection - Motivation

• Recall that source can be either a periodic
• impulse train spaced by F0 or random noise
• Autocorrelation function of a speech frame

• If x(n) is periodic in N, then R(k) is also
  periodic in N
• Thus, we can compute R(k) and check if its
  periodic

22
Pitch Detection Motivation

• The autocorrelation representation retains  too
  much of the information in the speech signal.We
  process the speech signal to make the periodicity
  more  prominent  while suppressing other
  distracting features of the signal.
• Techniques which perform this type of operation
  on signal are called  "spectrum flatteners" since
  they remove  the effects of the vocal ract 
  transfer function

23
Pitch Detection Motivation

• First we clip the frame using 3-level center
• clipping function

24
Pitch Detection Motivation

• Next we compute the modified autocorrelation
  function
• where can have only 3 different values

25
Pitch Detection-Algorithm
26
LPC Synthesizer