/EMBED

Efficient Methods for BEamforming Deconvolution

Primary LanguageMATLABOtherNOASSERTION

EMBED :: Efficient Methods for BEamforming Deconvolution.

EMBED is a small MATLAB package accompanying the journal paper:

Oliver Lylloff, Efrén Fernández-Grande, Finn Agerkvist, Jørgen Hald, Elisabet Tiana Roig and Martin S. Andersen. "Improving the efficiency of deconvolution algorithms for sound source localization." J. Acoust. Soc. Am. 138, 172 (2015); http://dx.doi.org/10.1121/1.4922516

The package contains two folders:

src/

  • FFTNNLS.m : FFT-NNLS algorithm
  • FISTA.m : FISTA algorithm
  • nnls.m : nnls problem f(x) = 1/2*||Ax-b||_2^2
  • zeropad.m : zeropads arrays before deconvolution procedure

example/

  • beamformer_setup.m : setup of grid size, opening angle, SNR
  • psf.m : calculates point-spread function
  • deconv.m : setup of deconvolution algorithms with number of iterations and zeropadding
  • micarray.mat : coordinates (x,y) of microphone array
  • run_example.m : run example and plot results

Usage: Add src/ to path and run run_example.m