基于GA理论的Volterra核辨识的多音激励设计

摘要

The strong peaks would be present in the design of multitone excitation signal of Volterra kernel identification for improperly choosing the phases of frequency components of the multitone signal, to overcome this problem, a novel design method of multitone excitation signal is proposed based on genetic algorithm (GA). Firstly, the output properties of Volterra kernels excited by multitone excitation signal are analyzed, and the conditions required by the design of multitone excitation of Volterra kernels identification are proposed. Secondly, minimizing the crest factor (CF) is considered as objective function, and the problem of choosing phases of multitone excitation signal is considered as a multiple variables optimized problem which can be solved by genetic algorithm. Finally, the values range of the variables, encoded mode and control parameters of GA are estimated in terms of practical problems. Theoretical analysis and experimental results indicate that the proposed method does not change the power spectrum of multitone excitation signal and can observably reduce CF. The proposed method avoids generating strong peaks in the multitone excitation signal and is a practical method.%为了克服以往在Volterra核辨识的多音激励信号设计过程中,由于不恰当的选择多音激励信号相位而导致出现信号出现的强波峰,提出了一种新的基于遗传算法(genetic algorithm,GA)的多音激励信号设计方法.分析了多音激励下Volterra核的输出性质,提出了Volterra核辨识多音激励信号设计应满足的条件；以最小化波峰因子(Crest factor,CF)作为目标函数,将多音激励信号的相位选择问题看成一个多元优化问题,便于采用遗传算法来求解该问题；根据实际问题确定变量的取值范围,编码方式及遗传算法的控制参数等.理论分析及实验结果表明,该方法在不改变功率谱的情况下,可显著降低信号的CF,避免了强波峰的产生,是一种实用的方法.