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.
Document Details
- Document Type
- Technical Report
- Publication Date
- Apr 01, 2010
- Accession Number
- ADA519885
Entities
People
- Alfredo V. Irizarry
Organizations
- Air Force Research Laboratory