Hi, I was just wondering if there is anyway to extract some SNR information without much knowledge about the signal or the noise. Let's say if I received some noisy measurement, is there anyway I can remove and filter out some of the noise? Thanks, --Rudy _____________________________ Posted through www.DSPRelated.com
SNR of an unknown signal
Started by ●January 27, 2014
Reply by ●January 27, 20142014-01-27
"rudykeram" <51467@dsprelated> writes:> Hi, > I was just wondering if there is anyway to extract some SNR information > without much knowledge about the signal or the noise. Let's say if I > received some noisy measurement, is there anyway I can remove and filter > out some of the noise?Without much knowledge, not much. -- Randy Yates Digital Signal Labs http://www.digitalsignallabs.com
Reply by ●January 27, 20142014-01-27
>"rudykeram" <51467@dsprelated> writes: > >> Hi, >> I was just wondering if there is anyway to extract some SNR information >> without much knowledge about the signal or the noise. Let's say if I >> received some noisy measurement, is there anyway I can remove andfilter>> out some of the noise? > >Without much knowledge, not much. >-- >Randy Yates >Digital Signal Labs >http://www.digitalsignallabs.com >Let's say the most I know is that it is a white noise, but not much info is available about its level or strength. Then is it yet not possible to get some accurate information about the signal? --Rudy _____________________________ Posted through www.DSPRelated.com
Reply by ●January 27, 20142014-01-27
On Tuesday, January 28, 2014 7:31:09 AM UTC+13, rudykeram wrote:> Hi, > > I was just wondering if there is anyway to extract some SNR information > > without much knowledge about the signal or the noise. Let's say if I > > received some noisy measurement, is there anyway I can remove and filter > > out some of the noise? > > > > Thanks, > > --Rudy > > > > _____________________________ > > Posted through www.DSPRelated.comWell, you could either make an educated guess that it was say white noise and then go for some form of adaptive filter. For example IEEE Xplore - A self-tuning filter for fixed-lag smoothing, by Hagander and Witenmark or some other form of blind signal estimation method.
Reply by ●January 27, 20142014-01-27
You can try to do some sort of spectral extrapolation, meaning first you can try to figure out what the white noise density is. Then you try and figure out where the signal is, and compare the in-band energy of the observation versus the expected in-band energy of the white noise component. Subtract the two and you get some rough estimate. But in the absence of a model or some sort of knowledge of what the hell signal you are trying to extract, or estimate, or model, I suspect that "SNR" is meaningless at best, misleading at worst.
Reply by ●January 27, 20142014-01-27
On Mon, 27 Jan 2014 12:31:09 -0600, rudykeram wrote:> Hi, > I was just wondering if there is anyway to extract some SNR information > without much knowledge about the signal or the noise. Let's say if I > received some noisy measurement, is there anyway I can remove and filter > out some of the noise?In a word -- no. With no knowledge of the signal or the noise characteristics, you have no way to distinguish between the signal and the noise. In that case, it becomes impossible to design a filter. Noise, after all, is signal, too -- it's just signal that carries information about processes that you don't care about. The best example of this that I can think of is if we stand next to each other, yakking on CDMA cell phones. I'm transmitting a spread-spectrum signal, you're transmitting a spread-spectrum signal, and there are two spread-spectrum signals coming back from the cell tower(s). My transmit signal is just noise to the part of the cell tower's algorithm that's decoding your transmit signal, your transmit signal is the same to my cell tower's algorithm, my cell tower's transmit signal is noise to you, and your cell tower's transmit signal is noise to me. Yet all four signals are valid signals to the people and devices who care about them (those people being you, me, whoever we're talking to, and every spy agency in the world, presumably). -- Tim Wescott Wescott Design Services http://www.wescottdesign.com
Reply by ●January 27, 20142014-01-27
On Mon, 27 Jan 2014 12:48:23 -0600, rudykeram wrote:>>"rudykeram" <51467@dsprelated> writes: >> >>> Hi, >>> I was just wondering if there is anyway to extract some SNR >>> information without much knowledge about the signal or the noise. >>> Let's say if I received some noisy measurement, is there anyway I can >>> remove and > filter >>> out some of the noise? >> >>Without much knowledge, not much. >>-- >>Randy Yates Digital Signal Labs http://www.digitalsignallabs.com >> >> > Let's say the most I know is that it is a white noise, but not much info > is available about its level or strength. Then is it yet not possible to > get some accurate information about the signal?Which signal? The spread-spectrum video carrying political discussions, the Super Bowl, porn, or some other socially irredeemable content? Or the detailed information about the life of the electrons in the various resistive materials in the area? -- Tim Wescott Wescott Design Services http://www.wescottdesign.com
Reply by ●January 27, 20142014-01-27
>On Mon, 27 Jan 2014 12:48:23 -0600, rudykeram wrote: > >>>"rudykeram" <51467@dsprelated> writes: >>> >>>> Hi, >>>> I was just wondering if there is anyway to extract some SNR >>>> information without much knowledge about the signal or the noise. >>>> Let's say if I received some noisy measurement, is there anyway I can >>>> remove and >> filter >>>> out some of the noise? >>> >>>Without much knowledge, not much. >>>-- >>>Randy Yates Digital Signal Labs http://www.digitalsignallabs.com >>> >>> >> Let's say the most I know is that it is a white noise, but not muchinfo>> is available about its level or strength. Then is it yet not possibleto>> get some accurate information about the signal? > >Which signal? The spread-spectrum video carrying political discussions, >the Super Bowl, porn, or some other socially irredeemable content? Or >the detailed information about the life of the electrons in the various >resistive materials in the area? > >-- > >Tim Wescott >Wescott Design Services >http://www.wescottdesign.com > >I see the point. Makes sense. Thanks -- Rudy _____________________________ Posted through www.DSPRelated.com
Reply by ●January 27, 20142014-01-27
rudykeram <51467@dsprelated> wrote:> I was just wondering if there is anyway to extract some SNR information > without much knowledge about the signal or the noise. Let's say if I > received some noisy measurement, is there anyway I can remove and filter > out some of the noise?There is much work on such for audio systems, called single-ended, though they don't work as well as double-ended ones that preprocess the signal. (The latter you might know as dbx, Dolby-B, Dolby-C.) Well, you have to know something generally about the signal, but not necessarily the specific signal. Audio signals are usually (maybe not in all cases) significantly different from the usual (not so white) noise. The two I know (a little) about are DNL and DNR. DNR has enough similarity to a Dolby-B decoder that it can be used on Dolby-B encoded source, in addition to uncoded sources. Other than for audio, I don't know of any, but the same ideas might apply. -- glen
Reply by ●January 28, 20142014-01-28
On Mon, 27 Jan 2014 13:36:05 -0600, Tim Wescott <tim@seemywebsite.really> wrote: [Snipped by Lyons]> >Which signal? The spread-spectrum video carrying political discussions, >the Super Bowl, porn, or some other socially irredeemable content? Or >the detailed information about the life of the electrons in the various >resistive materials in the area?Tim, Did you forget to take your medication this morning? Just jokin' Tim. [-Rick-]
