2
$\begingroup$

In ECE-700 Filterbank Notes the partitioning is $ \Delta = \frac{2\pi}{M}$ with center-frequencies at $ \omega_k = \frac{2\pi}{M}k$ which makes me wonder, considering that frequencies in the range from $\pi$ to $2\pi$ are just the negative versions of the range from $0$ to $\pi$, which divide into low- and high-pass frequencies.

In the follow-up picture the partitioning then stretches from $0$ to $\pi$ - is there someone so nice as to explain to me what I'm not understanding here?

$\endgroup$
1
$\begingroup$

The notes made me wonder as well, and I cannot suspect the author made a mistake. The first graph with periodicity is generic. The second one corresponds to dyadic iteration (wavelets), often displayed in $[0,\pi]$, except when dealing with complex wavelets where $[-\pi,\pi]$ is used. I agree, $[0,2\pi]$ could have been used instead, but I have never seen the $2^k$ partitioning mirrored at the end.

You can see from the rest of the paper that the author uses $[0,\pi]$.

$\endgroup$
1
$\begingroup$

It would appear that the author is only showing [0,π] for simplicity. This is often done in textbooks on the subject as well.

A full implementation would indeed have filters with centers that fill the range of [0,2π] or [-π, π], however you wish to visualize the spectrum. You are correct that for real signals, the negative frequencies are just a mirror of the positives; this however is not the case for a complex valued input signal, where the entire interval of [-π, π] can be unique. Nevertheless, for a near perfect reconstruction filter bank, our filters should cover the entire spectrum, such that no information is lost.

Summary: [0,π] is shown for simplicity. Realistically, the filters cover the full [0, 2π] swath.

$\endgroup$

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.