A High Precision Parameter Estimation Method of Double Exponential Pulse Based on Artificial Neural Network

Double exponential pulse is the waveform of high-power electromagnetic environments. 10-90% rise time
$t_{r}$
and pulse width at half maximum
$t_{w}$
are the important parameters of the waveform. They can be determined by
$\alpha$
and
$\beta$
which are the key parameters of double exponential function. The transformation from
$t_{r},\ t_{w}$
to
$\alpha,\ \beta$
is frequently used and very necessary in EM simulation and data analysis. In this paper, based on artificial neural network we create a new training data set and establish a simple architecture.
$\alpha,\ \beta$
can be given out directly from the model. Compared to piecewise linear fitting and look-up table method, the method proposed in this paper has the advantages of both high precision and small amount of pre-processing data. The maximum estimation error of both
$\alpha$
and
$\beta$
is less than 0.1%.