AAC | ADPCM | Convolution | DAFx | FFT | IIR | Mixer | MP3 | MPEG | MPEG-4

Hi,

I am using Linear prediction coding technique for my speach synthesizer
here is the process i am using.

I pass my speech samples to autocorrelation function to get 
autocorrelation. then this autocorrelation sequence is passed to 
levinson durbin recursion method to obtain filter parameters or filte 
coefficients. then these coefficents are used to obtain the original 
speech back in the following way.

S= Array of samples
L= Levinson Durbin/filter coefficients
if i want to get sample six from first five samples then
S6 = (S1 * L1) + (S2 * L2) + (S3 * L3) + (S4 * L4) + (S5* L5)

I dont know my procedure is correct or not. The answer i am getting is 
not correct so please guide me about the procedure.

Regards
Imran Farooq

(You need to be a member of audiodsp -- send a blank email to audiodsp-subscribe@yahoogroups.com )

Hi,
   
     Linear predictions is not exactly interpolation. In this case it's work is as follows. 
    Linear prediction filter predicts the spectrum envelope of the signal. This is subtracted
from the original signal. Now the remained signal called as residual signal contains only the
pitch information. This pitch information is removed from the residual signal and is encoded
using adaptive and fixed code books. The pitch information and adaptive and fixed code book
indices are transmitted to the decoder. In the decoder side using the predictor filter
coefficients, synthesis filter is constructed. The excitation signal is obtained from the
adaptive and fixed code book indices. Pitch content is added to the output of the codebooks.
This is the excitation signal to the sythesis filter. In this way the linear prediction filter
is useful in the speech coder. 
   
     if you have any further queries, mail me back
   
  Regards
   sridhar.a

mastershaikh <m...@yahoo.com> wrote:
          Hi,

I am using Linear prediction coding technique for my speach synthesizer
here is the process i am using.

I pass my speech samples to autocorrelation function to get 
autocorrelation. then this autocorrelation sequence is passed to 
levinson durbin recursion method to obtain filter parameters or filte 
coefficients. then these coefficents are used to obtain the original 
speech back in the following way.

S= Array of samples
L= Levinson Durbin/filter coefficients

if i want to get sample six from first five samples then
S6 = (S1 * L1) + (S2 * L2) + (S3 * L3) + (S4 * L4) + (S5* L5)

I dont know my procedure is correct or not. The answer i am getting is 
not correct so please guide me about the procedure.

Regards
Imran Farooq

(You need to be a member of audiodsp -- send a blank email to audiodsp-subscribe@yahoogroups.com )