中文摘要：

在经典的基因表达式编程算法（GEP）理论基础上，提出了一种固定结构多种群GEP算法（FSMGEP）。该算法在解的描述中采用了具有固定长度的线性符号串结构，通过重新定义遗传算子及适应度计算过程降低了算法的计算复杂性；在求解过程中采用多种群协同进化思想增强解的多样性，并引入爬山算法对参数进行局部优化，提高了算法的求解精度与效率。最后在语音信号序列预测中的应用表明，FSM GEP算法较经典GEP在收敛速度和收敛精度上均有明显提高。

英文摘要：

A fixed structure and multi-population GEP algorithm （FSMGEP） was proposed based on the tradition-al gene expression programming （GEP） theory .The linear string structure with fixed length was used in the de-scription of the individual ,which decreased the computation complexity .The collaborative evolutionary thought of multi-population was introduced in the FSMGEP to improve the diversity of the individuals .And the hill-climbing algorithm was utilized to optimize the parameters ,which improved the accuracy and efficiency of the FSMGEP .The application in the prediction of speech signals showed that the FSMGEP has better convergence speed and precision than the traditional GEP .