Hi Can anybody help me to clear up question about what should be called parametric methods. For example, in some sources the Capon's beamformer (which is also called MVDR and Frost beamformer) is considered to be spectral method in others it is a model-based parametric beamformer (or DOA estimator). The same is about MUSIC. Another question is that if maximum likelyhood is a model-based approach, why it is considered as similar (or completely the same) to minimum variance criterion, which is used for nonparametric beamformers. It would be very helpful if you give some references where this issue is clearly described. Thanks in advance Best regards This message was sent using the Comp.DSP web interface on www.DSPRelated.com
Parametric Methods
Started by ●October 14, 2005
Reply by ●October 14, 20052005-10-14
sotnik wrote:> Hi Can anybody help me to clear up question about what should be called > parametric methods. For example, in some sources the Capon's beamformer > (which is also called MVDR and Frost beamformer) is considered to be > spectral method in others it is a model-based parametric beamformer (or > DOA estimator).When you derive the MVDR bearmformer, the direction constraint is assumed to be actually pointed at source you want. In practice, this isn't usually the case, so you have a parameter(s) that you need to estimate. One can estimate the parameter or do a number of other things like diagonal loading. The same is about MUSIC. Another question is that if> maximum likelyhood is a model-based approach, why it is considered as > similar (or completely the same) to minimum variance criterion, which is > used for nonparametric beamformers.Even non-parametric techniques have modeling assumptions. Sometimes you need the parameter and sometimes you don't. One of the interesting things about the theory is that you often get similar solutions for problems that are derived in different ways. It doesn't always happen that way. It would be very helpful if you give> some references where this issue is clearly described. > > Thanks in advance > Best regards > > > > This message was sent using the Comp.DSP web interface on > www.DSPRelated.com
Reply by ●October 15, 20052005-10-15
How about Van Trees Volume 4. Detection, Estimation, and Modulation Theory, Part IV, Optimum Array Processing In article <dio734$r05$1@newslocal.mitre.org>, Stan Pawlukiewicz <spam@spam.mitre.org> wrote:>sotnik wrote: >> Hi Can anybody help me to clear up question about what should be called >> parametric methods. For example, in some sources the Capon's beamformer >> (which is also called MVDR and Frost beamformer) is considered to be >> spectral method in others it is a model-based parametric beamformer (or >> DOA estimator). > >When you derive the MVDR bearmformer, the direction constraint is >assumed to be actually pointed at source you want. In practice, this >isn't usually the case, so you have a parameter(s) that you need to >estimate. One can estimate the parameter or do a number of other things >like diagonal loading. > > > >The same is about MUSIC. Another question is that if >> maximum likelyhood is a model-based approach, why it is considered as >> similar (or completely the same) to minimum variance criterion, which is >> used for nonparametric beamformers. > >Even non-parametric techniques have modeling assumptions. Sometimes you >need the parameter and sometimes you don't. > >One of the interesting things about the theory is that you often get >similar solutions for problems that are derived in different ways. It >doesn't always happen that way. > >It would be very helpful if you give >> some references where this issue is clearly described. >> >> Thanks in advance >> Best regards >> >> >> >> This message was sent using the Comp.DSP web interface on >> www.DSPRelated.com
Reply by ●October 17, 20052005-10-17
John Herman wrote:> How about Van Trees Volume 4. > > Detection, Estimation, and Modulation Theory, Part IV, Optimum Array > ProcessingWhat about it?> > > In article <dio734$r05$1@newslocal.mitre.org>, Stan Pawlukiewicz > <spam@spam.mitre.org> wrote: > >>sotnik wrote: >> >>>Hi Can anybody help me to clear up question about what should be called >>>parametric methods. For example, in some sources the Capon's beamformer >>>(which is also called MVDR and Frost beamformer) is considered to be >>>spectral method in others it is a model-based parametric beamformer (or >>>DOA estimator). >> >>When you derive the MVDR bearmformer, the direction constraint is >>assumed to be actually pointed at source you want. In practice, this >>isn't usually the case, so you have a parameter(s) that you need to >>estimate. One can estimate the parameter or do a number of other things >>like diagonal loading. >> >> >> >>The same is about MUSIC. Another question is that if >> >>>maximum likelyhood is a model-based approach, why it is considered as >>>similar (or completely the same) to minimum variance criterion, which is >>>used for nonparametric beamformers. >> >>Even non-parametric techniques have modeling assumptions. Sometimes you >>need the parameter and sometimes you don't. >> >>One of the interesting things about the theory is that you often get >>similar solutions for problems that are derived in different ways. It >>doesn't always happen that way. >> >>It would be very helpful if you give >> >>>some references where this issue is clearly described. >>> >>>Thanks in advance >>>Best regards >>> >>> >>> >>>This message was sent using the Comp.DSP web interface on >>>www.DSPRelated.com