Hi, I am trying to implement a time series version of the popular MVDR beamformer. As I understand, the data for the beamformer is spatial samples of a plane wave and the antenna array is steered through a range of angles to get a peak which indicates the DOA and thereby helps in estimating the source complex amplitude. I was wondering if similar concept could be applied if my data now is a window worth of temporal samples of a noisy sinusoid. This would more sound like 'bandforming' because we are now trying to separate source from noise using frequency components. I am unclear as to how to model such a scenario. Any help would be greatly appreciated. best, Abhi
Time series version of MVDR beamforming
Started by ●September 9, 2008
Reply by ●September 10, 20082008-09-10
On Sep 9, 7:35 pm, "Abhimanyu" <sowmya.lo...@gmail.com> wrote:> Hi, > > I am trying to implement a time series version of the popular MVDR > beamformer. As I understand, the data for the beamformer is spatial samples > of a plane wave and the antenna array is steered through a range of angles > to get a peak which indicates the DOA and thereby helps in estimating the > source complex amplitude. I was wondering if similar concept could be > applied if my data now is a window worth of temporal samples of a noisy > sinusoid. This would more sound like 'bandforming' because we are now > trying to separate source from noise using frequency components. I am > unclear as to how to model such a scenario. Any help would be greatly > appreciated. > > best, > AbhiSounds like you want to do line frequency estimation, is this correct? It's actually the prototype problem, which solution was then re-used to solve the MVDR problem. The best reference is Stoica's book on Spectral Estimation, or read up on any "line spectrum estimation" papers.
Reply by ●September 12, 20082008-09-12
On Sep 10, 4:12�pm, julius <juli...@gmail.com> wrote:> On Sep 9, 7:35 pm, "Abhimanyu" <sowmya.lo...@gmail.com> wrote: > > > Hi, > > > I am trying to implement a time series version of the popular MVDR > > beamformer. As I understand, the data for the beamformer is spatial samples > > of a plane wave and the antenna array is steered through a range of angles > > to get a peak which indicates the DOA and thereby helps in estimating the > > source complex amplitude. I was wondering if similar concept could be > > applied if my data now is a window worth of temporal samples of a noisy > > sinusoid. This would more sound like 'bandforming' because we are now > > trying to separate source from noise using frequency components. I am > > unclear as to how to model such a scenario. Any help would be greatly > > appreciated. > > > best, > > Abhi > > Sounds like you want to do line frequency estimation, is this correct? > It's actually the prototype problem, which solution was then re-used > to > solve the MVDR problem. > > The best reference is Stoica's book on Spectral Estimation, or read > up on any "line spectrum estimation" papers.If I recall correctly, the formulation of the MVDR beamformer is analogous to that of the LCMV (linearly-constrained, minimum-variance) filter, which I believe is what you want. Try searching for that topic also. Jason