You can use \smash and a \vphantom to make it work.
\documentclass{article}
\usepackage{mathtools}
\begin{document}
\[\sqrt{\vphantom{\int_{\text{-}\infty}^{t}x}\smash{
\underbrace{
\frac{1}{\tau}
\underbrace{
\int_{\text{-}\infty}^{t}
\underbrace{
p_{X}^{\,2}\left(\xi\right)\cdot e^{\frac{\text{-}\left(t\text{-}\xi\right)}{\tau}}
}_{\substack{\text{Expontial}\\\text{time-weighting}\\\text{of $p_{X}^{\,2}\left(\xi\right)$}}} \,\mathrm{d}\xi
}_{\text{Exponential integration}}
}_{\substack{\text{Exponential averaging}\\\text{= “Quasi mean-square”}}}
}}\]
\end{document}
The output looks like this:

I should note that this does not appear to work in inline math-mode. The size of the vphantom does not correctly reflect the integral sign then. Perhaps someone else can explain why this is the case, since I don't know.
Edit after @egreg's comment, an additional vphantom should be added, to ensure the displaymath gets the height of the braces as well. The modified code would look like this:
\documentclass{article}
\usepackage{mathtools}
\begin{document}
\[\sqrt{\vphantom{\int_{\text{-}\infty}^{t}x}\smash{
\underbrace{
\frac{1}{\tau}
\underbrace{
\int_{\text{-}\infty}^{t}
\underbrace{
p_{X}^{\,2}\left(\xi\right)\cdot e^{\frac{\text{-}\left(t\text{-}\xi\right)}{\tau}}
}_{\substack{\text{Expontial}\\\text{time-weighting}\\\text{of $p_{X}^{\,2}\left(\xi\right)$}}} \,\mathrm{d}\xi
}_{\text{Exponential integration}}
}_{\substack{\text{Exponential averaging}\\\text{= “Quasi mean-square”}}}
}}
\vphantom{\underbrace{
\frac{1}{\tau}
\underbrace{
\int_{\text{-}\infty}^{t}
\underbrace{
p_{X}^{\,2}\left(\xi\right)\cdot e^{\frac{\text{-}\left(t\text{-}\xi\right)}{\tau}}
}_{\substack{\text{Expontial}\\\text{time-weighting}\\\text{of $p_{X}^{\,2}\left(\xi\right)$}}} \,\mathrm{d}\xi
}_{\text{Exponential integration}}
}_{\substack{\text{Exponential averaging}\\\text{= “Quasi mean-square”}}}
}
\]
A long sentence after it to see if there are problems with overruning.
\end{document}
And its output like this:
