中文摘要：

根据非稳态超声信号的特点,提出一种改进的新型人工鱼群智能优化的稀疏分解算法,构造了人工鱼群搜索稀疏分解最佳原子的优化模型 利用人工鱼群方法并行寻优、全局收敛性好的特点,搜索最优原子,降低了稀疏分解匹配追踪算法的复杂度,减少了传统匹配追踪中超完备字典对存储空间的占用 针对鱼群搜索特点,对初始鱼群分布及鱼群行为进行改进,解决了原始算法鱼群初始覆盖空间的不确定性,改进后的聚群和追尾行为有效地提高了鱼群算法的收敛速度且算法稳定 实验结果表明,将改进后的算法用于超声缺陷信号的提取时,与小波方法相比较,信号的质量和性能指标均有显著改善。

英文摘要：

According to the characteristics of non-stationary ultrasonic flaw signals, a new fast optimized sparse decomposition algorithm is proposed based on improved artificial fish swarm in this paper, and an optimization model of searching atoms for artificial fish swarm algorithm is built. Arificial fish swarm has the advantages of distributed parallel searching ability, good global astringency and is employed to search the best matching atoms, which reduced the complexity of sparse decomposition and decreased the occupation of memory space that is needed by over-complete dictionary in traditional matching pursuit. According to the features of fish swarm searching, the initial value and behavior of artificial fish swarm are improved, which solve the uncertain initial space covered by fish swarm. The convergence speed of the artificial fish swarm algorithm is increased through improving the behavior of artificial fish. The proposed algorithm is stable compared with original algorithm. Experiment results show that the signal quality and performance parameters are improved obviously compared with wavelet transform signal processor method.