Low Complexity Search Method for G.723.1 MP-MLQ Algorithm

The Multi-pulse Maximum Likelihood Quantization (MP-MLQ) algorithm with high complexity and high quality in its analysis-by-synthesis optimization has been used by many speech coding standards. To reduce the computational complexity of MP-MLQ in ITU-T G.723.1 speech codec, in this paper, we propose a fast search algorithm by using a designed function and the multi-track positions structure to predict the candidate pulses for each subframe. Simulation results show that the degradation of PESQ score for the pro-posed approach is about 0.069 relative to the original search procedure. However, the proposed approach reduced the computational complexity about 98.346% with perceptually negligible degradation.