基于改进匹配追踪算法的语音信号处理研究∗

摘要

针对匹配追踪算法最匹配原子的搜索存在计算量较大、时间过长、效率不高等问题，根据粒子群算法的搜索特性，使用概率选择方式更新全局极值，利用混沌映射改善随机值，并将改进后的粒子群算法融入到标准匹配追踪算法中，通过重构语音信号，证明改进后的算法具有一定的可行性，其重构质量和运行速度均优于标准匹配追踪算法。%Aiming at large calculation, time-consuming and low efficiency for the searching the best atomic in matching pursuit algo￣rithm, and based on the searching characteristic of PSO (Particle Swarm Optimization), the update of globalextremum is done with probability selection method, and the random value improved by Chaos mapping. Meanwhile, the modified PSO is integrated into the standard matching pursuit algorithm. Reconstruction of speech signal indicates that the modified algorithm is of certain feasibility, and both the reconstruction quality and operation speed are superior to the standard matching pursuit algorithm.