What does linear prediction assume about the speech signal s(n)?
It is represented by an auto-regressive model
The model includes an excitation signal x(n) and p linear prediction coefficients.
What are the components of the auto-regressive model for speech signal s(n)?
- G
- x(n)
- L
- ak (k=1 to p-1)
The values of G and ak are efficiently solved by the Levinson-Durbin recursion.
What does the Z transform give for the speech signal?
S(z) = G / (1 - E = LaR = k)
This equation relates the Z transform to the filter’s frequency response.
What is the formula for the cepstrum of the speech signal?
Cs(T) = F{In S(w)}
This definition is crucial for understanding the cepstral coefficients.
How are the cepstral coefficients modeled in the vocal-tract filter?
Using the low lag coefficients, 0 ≤ T < p
This modeling is essential for analyzing the spectral envelope.
What is the relationship between the cepstral coefficients and the filter’s frequency response?
|H(w)| = exp(F{ch(T)})
This equation shows how cepstral coefficients relate to the frequency response of the filter.
What type of model is used in linear prediction?
Auto-regressive (AR) model
This model is fundamental in speech signal processing.
