I am a college teacher specializing in signal processing. One of the graduate courses that I teach is "Adaptive Filter Theory". In that class, one of the most frequently asked questions is: "If we already have access to the desired response, then why we bother with filtering? If we have the information, in advance, as to what the output of the filter should be (desired response), then just use that, instead of trying to get it from a noise corrupted and degraded input signal.". I have my own answer to this question. Nevertheless, I would like to read what adaptive filtering experts have to say on that seeming dilemma. I am looking forward to stimulating comments.
You need the desired signal just to benchmark the different filters according to signal characteristics
Why we bother filtering ????!!!! ok will start with applications :
1. Signal conditioning (Measurement using transducers) : Clean/ prepare analog/physical world signals for next level of processing- Clean to remove noise form device (starting with resistors) or unwanted signals
Prepare to check for ranges at which circuit efficiency is maximum (benefit of doubt using 3dB or optical fiber -6dB)
2. Communications: Wide band of frequency as they travel common thing is pulse spreading leading to interference /distortions /reduction in energy , so amplify (noise gets added ) => filter .
Next , carrier removal (because of modulation or to increase the BW) , use filters for particular frequency BW (pre-emphasis and de-emphasis ) . Super-Hetrodyne receivers (for a range of frequency ) , upconverters (uplinks) , down converters (down links ) , the list goes on and on and on
Now , why adaptive typical application in RADAR - has a range of frequencies and reflections has a subset , so it is inevitable . Implemented using DSPs hence largely digital filters are programmable and tunable or rather reduce computational complexity to some extent .
We have information but the quality say for next level of processing, it is necessary to prepare , for maximizing and getting clarity of signals that have traveled (media or without) . Eg: Echo cancelers in telephone or ear implants (to a certain level) are some sort of Chebyshev filter .
The question :Is there any improvement in instrumentation or device that reduce noise levels in frequency range or do not corrupt ???
Once we do that (if possible ) we do not need filters adaptive or not ....
It's all about noise... You generally don't have the desired signal by itself... often all you have is a noise-corrupted version of it... You are relying on optimizing the correlation. The filter attempts to predicts a noise-free version of the signal.
- Badges
- Science topic
- Similar topics
- Computer Science and Engineering
- Artificial Intelligence