The Spectral Mixture Models: A Minimum Information Divergence Approach

Abstract

The objective of developing the Spectral Mixture Model Algorithm was to provide some intelligent algorithm that could be utilized for spectral sensing in wideband receivers. The methodology was discussed initially in report AFRL-RI-RS-TR-2008-266. The current report is a refinement of the technique with the objective of presenting the concept to a broader audience. The Spectral Mixture is a generalization of the Expectation Maximization algorithm. The algorithm reduces the information divergence of two distributions by adjusting its parameters. The algorithm can be applied to histogram data or sample points for signal decomposition of multimodal signal in terms of mixture elements. The model was applied to spectral analysis with good success in the one dimensional case. To achieve better convergence, the algorithm may require the constraint of some of the parameters by imposing boundary conditions or preventing changes. This research explored some potential applications of the algorithm. These include: spectral characterization, speech compression, deconvolution and image processing. The results are summarized in this report.

Open PDF

Document Details

Document Type
Technical Report
Publication Date
Apr 01, 2010
Accession Number
ADA519885

Entities

People

  • Alfredo V. Irizarry

Organizations

  • Air Force Research Laboratory

Tags

Communities of Interest

  • Materials and Manufacturing Processes

DTIC Thesaurus Topics

  • Air Force
  • Air Force Research Laboratories
  • Algorithms
  • Bandwidth
  • Compression
  • Convergence
  • Data Sets
  • Equations
  • Frequency
  • Government Procurement
  • Governments
  • Histograms
  • Image Processing
  • Information Exchange
  • Information Processing
  • Probability
  • Speech Compression

Readers

  • Finite Element Method (FEM) for solving Partial Differential Equations (PDEs)
  • Operations Research
  • Radar Systems Engineering.