Cisco Support Community



Algebraic Code Excited Linear Prediction (ACELP)

Complete  Definition

Algebraic Code Excited Linear Prediction (ACELP) is a speech encoding   algorithm where a limited set of pulses is distributed as excitation to linear   prediction filter.

The ACELP method is widely employed in current speech coding standards such   as AMR, EFR, AMR-WB and ITU-T G-series standards G.729 and G.723.1.

ACELP is a registered trademark of VoiceAge Corporation in Canada and/or   other countries and was developed in 1989 by the Université de Sherbrooke in   Canada.

ACELP is based on CELP coding  model, but ACELP codebooks have a  specific algebraic structure imposed upon  them.

A 16-bit algebraic codebook shall be used in the innovative codebook search,   the aim of which is to find the best innovation and gain parameters. The   innovation vector contains, at most, four non-zero pulses.

In ACELP a block of N speech samples is synthesized by filtering an   appropriate innovation sequence from a codebook, scaled by a gain factor g c,   through two time varying filters.

The long-term or pitch, synthesis filter is given by:

    1/B(z) = 1/(1 - gpz-T)

The short-term synthesis filter is given by:

     H(z) = 1/A(z) = 1/(1 + ZIGMAi = 1 to Paiz-i)

ACELP  Wikipedia Definition

Also See