基于三阶Rényi熵的核参数优化

Optimization of Kernel Parameters Based on Third-Order Rényi Entropy

在Cohen类时频分布中,为使减小交叉项与保持高的时频聚集性二者之间达到最佳折中,提出了一种基于三阶Rényi熵的核参数优化算法.根据三阶Rényi熵对交叉项的近似不变性,通过搜索三阶Rényi熵随核参数变化曲线下降由快变慢的转折点,可以获得最优核参数.理论分析和仿真结果表明:根据三阶Rényi熵对核参数进行优化,可以使核函数与信号达到最佳匹配,从而得到高性能的时频分布.

In order to obtain an optimal tradeoff between cross-term reduction and high time-frequency concentration in time-frequency distribution of Cohen＇s class, an optimization algorithm of kernel parameters based on third-order Renyi entropy was proposed. From the asymptotic cross term invariance of third-order Renyi entropy, the optimal kernel parameters can be obtained by searching the transition of the curve of third-order Renyi entropy versus kernel parameters. The theoretical analysis and simulation results show that the optimization of kernel parameters based on third-order Renyi entropy can match the kernel function best with signals to yield a high-performance time- frequency distribution.