Forgive me is this is an ill-posed question.

Is there any such thing as a 'convolution theorem' for the cross-correlation.

Namely, the convolution theorem states that: $$ x[n] * h[n] = \sum_{k=0}^{\infty}h[k] x[n-k] = X(f)H(f) $$ Where $X(f)$ is the Fourier transform of $x(t)$.

Is there any equivalent theorem for cross-correlation, i.e.: $$ \textrm{corr}(x[n],h[n])=r_{xh}[k]= \sum_{k=0}^{\infty}h[k] x[n+k] = F_?[X(f),H(f)] $$

In other words, is there a way to express $r_{xh}[k]$ in terms of $X(f)$ and $H(f)$, or even in terms of the autocorrelation $r_x[k]$ and $r_h[k]$.

I tried self-deriving this but couldn't come up with anything.


  • $\begingroup$ Your cross-correlation is a function of $n$ (not $k$). $\endgroup$
    – Gilles
    Mar 29, 2016 at 22:00

2 Answers 2


Yes. Cross correlation is the same as convolution with a time reversed signal. That corresponds to conjugate komplex in the frequency domain.

$R(\omega) = X(\omega ) \cdot H(\omega )^*$


To stick to proper notations, if $\mathcal F \left\{x[n]\right\}=X(\omega)$ and $\mathcal F \left\{h[n]\right\}=H(\omega)$, what you mean is the Fourier transform of a convolution gives the product of the transforms of each term:

$$ \mathcal F \left\{x[n] * h[n]\right\} = \mathcal F\left\{\sum_{k=0}^{\infty}h[k] x[n-k]\right\} =X(\omega)\cdot H(\omega) $$

As already mentioned, there is a similar formulation for correlation, the correlation theorem. And again, the Fourier transform of the correlation: $$ \mathcal F\left\{\mathrm{corr}\left(x[n],h[n]\right)\right\}=\mathcal F\left\{r_{xh}[n]\right\}= \mathcal F\left\{\sum_{k=0}^{\infty}h[k] x[n+k] \right\}= X(\omega)\cdot H^*(\omega) $$

The product $X(\omega) H^*(\omega)$ is called cross-energy density spectrum and noted $S_{xh}(\omega)$, and this is just $\mathcal F\left\{r_{xh}[n]\right\}$. The $^*$ sign is for conjugate.

Note that your convolution/correlation are on discrete-time signals $x[n]$ and $h[n]$. So $X(\omega)$ and $H(\omega)$ are transforms of $x[n]$ and $h[n]$ respectively, and not of continuous-time signals $x(t)$ and $h(t)$.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.