指数遗忘分布性能分析

该文分析讨论了指数遗忘变换(EFT)的信号瞬时频率表示性能。EFT采用单边指数窗口对信号进行采样和加权，因而可以利用迭代运算提高计算效率。与其他时频表示方法相比，EFT在处理较大长度的数据时具有优势，且更易于利用硬件实现。该文对EFT的瞬时频率表示偏差、均方差值的统计特性与信噪比及遗忘系数的关系进行了分析，给出了相应的仿真结果。同时为克服单边指数窗口的缺陷，提出了采用对称窗口的双边指数遗忘分布的计算形式，该方法在保持原有指数遗忘分布计算效率较高的优点的同时能够大大减小瞬时频率的表示偏差。

Analysis of Exponentially Forgetting Distribution

The analysis of exponentially forgetting transform (EFT) is made in this paper. The application of a sliding single-sided exponential window, make it possible the iterative operation in EFT which greatly raises computation efficiency. Compared with other time-frequency distributions, EFT is especially useful in case of large data length and more suitable for hardware implementation. The relationships of bias and mean square error of EFT with SNR and forgetting coefficient are studied. And in order to overcome the drawbacks of single-sided window, a new time-frequency distribution using double-sided exponential window which is computation efficient with much bias reduction is introduced.