autoregressive model analysis

[Biruntha]

Hi,
can anyone briefly explain why we need to do autoregressive model analysis for audio files? give some introduction to autoregressive model analysis?

 

[SanjKrish]

AR or All pole model is done to compress your data.

The traditional method of storing audio is to store each of its samples at all the sampling intervals.

In AR modeling you model the signal as thought it is the output of a LTI system(causal and stable of-course) with white noise as its input.

The LTI system H(Z) = 1/(b0 + b1(z-1) + ... )

All you need is to determine the coeffs b0 .. bn and the variance of the white input x

Then using the relation y = var(x) |H(e^jw)| you can construct the output audio.

Thereby instead of storing the samples you just store b0.. bn and var(x) which is less data. This is how compression is achieved.

 

[Biruntha]

thankyou for your rly , can you suggest some usefull links about this?

 

[SanjKrish]

The book "Statistical signal processing" by Hayes gives a good idea of AR signal modelling.