摘要
研究过完备原子库信号分解优化算法问题,信号的稀疏表示在信号去噪、信号压缩等方面有明显的优势。但因稀疏分解计算量巨大需要较长的计算时间,难以满足实时性要求,在实际应用中受到极大的限制。为此提出人工蜂群算法,具有需要设置的参数少、收敛速度快、鲁棒性强等优点,快速寻找匹配跟踪过程中每一步的近似最佳原子,对信号进行有效地稀疏分解。改进传统迭代的终止条件,以克服传统的迭代终止条件难以选择合适迭代终止阈值的问题,实现信号快速稀疏分解。实验结果表明,改进算法对信号的稀疏分解质量与粒子群算法和遗传算法相当,但运算速度均优于粒子群算法和遗传算法。
Sparse representation of signals has a distinct advantage in signal denoising, signal compression, etc. Because sparse decomposition has an enormous computation amount and takes a long computing time, it is difficult to meet real-time requirements and severely limited in practice. Artificial Bee Colony （ABC） algorithm needs fewer parameters, converges fastly, has strong robustness, etc. It can quickly search for the best approximation of atoms to match the signals at every step of the process. The paper also improved the traditional iteration termination conditions to choose the appropriate terminate threshold. Experimental results show that the proposed algorithm is equal to parti- cle swarm algorithm and genetic algorithm in quality, but is better than particle swarm algorithm and genetic algorithm in speed.
出处
《计算机仿真》
CSCD
北大核心
2012年第11期247-250,共4页
Computer Simulation
基金
国家自然科学基金(69674012)
重庆市科技攻关计划(CSTC2009AC3037)
关键词
信号
稀疏分解
匹配追踪
人工蜂群算法
Signal
Sparse decomposition
Matching pursuit
Artificial bee colony algorithm （ABCA）
