Can you split an signal into smaller windows segments before applying filters/downsampling it? Basically I wish to do a sliding window to the signal, but as it is too huge, the filtering and downsampling takes too long to process. Hence, I am wondering if it is possible to slice the signal into the windows before applying the filters and downsampling it. Will this induce any issue?
A Hamming window is a good way to avoid edge artefacts.
You might also want to revisit the Nyquist criterion (for reconstruction after decomposition): low pass before sampling...
Careful with what you do there.
You might prefere a sliding window (in the time domain).
If you work on image processing, people segment an image into sub-images and perform transforms (like DCT) on the smaller images.
Break your filtering into multiple stages. If using large windows, chain together several layers of smaller windows & downsampling. The CIC/Hogenauer filter is ideally suited to this task (the CIC filter is essentially a cascaded moving average filter, doing the 'slicing' that you want by default), but other filters can be used as well, as long as you design your filter transfer function to avoid possible aliasing products in decimated samples. The CIC filter is a workhorse filter used in many software radio systems that should be in everyone's DSP toolbox! Have fun!
If the effective span of the filter is limited (FIR or fast decaying response), you can split your signal in windows extended at the extremities (so that they overlap). This way, you can compute all results from the exact input data.
- Badges
- Science topic
- Similar topics
- Engineering
- Electrical Engineering
- Signal Processing