An Independent Component Analysis Blind Beamformer
Abstract
Independent Component Analysis (ICA) has been proven to be a very successful method for separating mixed signals blindly. ICA works by using the assumption that signal mixtures are combinations of independent signals. Up to now, ICA has been primarily used to separate signals where multiple amplitude combinations of these signals exist. The most popular of these ICA methods is Blind Source Separation (BSS). This thesis will expand on this to use the theory of ICA to separate mixed signals that are mixed by a beamformer. A new algorithm will be developed that combines BSS and a new blind beamforming method to provide an estimate 6f the original unmixed signals and simultaneously learns their corresponding directions. This will all be performed without using any a priori knowledge about the source waveforms or their directions. Results from this algorithm were very promising and worked to separate multiple unknown signals that propagated from different unknown directions. Signal to Noise and Interference Ratios (SNIR) of the estimated signals, were found to significantly improve using this algorithm along with accurate estimates of their directions. BSS was also used in the algorithm to speed up convergence time and provide cleaner versions of the estimated signals. This algorithm was also shown to work well for wideband signals by using wideband network.
Document Details
- Document Type
- Technical Report
- Publication Date
- Dec 01, 2000
- Accession Number
- ADA384795
Entities
People
- Marc L. Salerno
Organizations
- Pennsylvania State University