Effective complexity reduction in codebook search for ACELP?

Algebraic Code Excited Linear Prediction (ACELP) speech coding algorithm provides high speech quality, but it also has high complexity, so it is difficult to implement in real-time system, and the codebook search is the main operational load. This paper proposes an effective algorithm, first, the discontinuous pitch estimation is proposed in adaptive codebook search, the time decimation factor is used in different delays, and then a modified pulse-replacement search algorithm is proposed in stochastic codebook search, the position prediction technique and judging mechanism are used to control the number of codebook search. Finally, the improved algorithm is simulated in G.729A, the result show that the proposed algorithms not only acquire complexity reduction but also get a high speech quality.