Abstract: The Time-Frequency Moving Average （TFMA） model algorithm which is a method of non-stationary signal processing  degenerate under [α] stable distribution environment, the fractional lower order statistics covariance  is introduced and the improved Fractional Lower Order Time-Frequency Moving Average algorithm （FLO-TFMA） model algorithm is proposed. The parameters estimation of FLO-TFMA model is developed and time-frequency spectrum estimation is given based on the FLO-TFMA model. By comparing the Mean Square Error （MSE） of parameter estimation and spectrum estimation of the TFMA model algorithm and the proposed FLO-TFMA model algorithm under [α] stable distribution environment condition, simulations show that the parameters estimation precision of the FLO-TFMA model algorithm is better than TFMA model algorithm, the TFMA model spectrum estimation can not work, and FLO-TFMA model algorithm provides better performance of time-frequency spectrum.

Key words: [α] stable distribution, fractional lower order statistic, moving average model, non-stationary process, time-frequency spectrum estimation

摘要： 针对稳定分布环境下非平稳过程分析方法时频滑动平均（TFMA）模型算法的退化，引入分数低阶统计量共变，提出了一种改进的分数低阶时频时频滑动平均（FLO-TFMA）模型算法。推导了FLO-TFMA模型的参数求解过程，给出了基于FLO-TFMA模型的时频谱估计。通过在稳定分布环境下对TFMA模型算法和所提出的FLO-TFMA模型算法的参数估计均方误差（MSE）比较和时频谱估计比较，仿真结果表明，FLO-TFMA模型算法的参数估计精度优于TFMA模型算法，TFMA模型时频谱估计完全失效，而FLO-TFMA模型时频谱算法能较好地进行时频谱估计。

关键词: 稳定分布, 分数低阶统计量, 滑动平均模型, 非平稳过程, 时频谱估计