Signal Processing in Subspaces.
Abstract
Reduced dimension or subspace signal processing algorithms are studied for several classes of signal processing problems. The approach consists of mapping data into a subspace with a rectangular matrix transformation prior to application of the signal processing algorithm. This approach reduces computational complexity, reduces the variability associated with quantities estimated from data, and generally introduces some asymptotic performance loss. However, in situations with limited data, the impact of reduced variance generally dominates the asymptotic performance loss and a net improvement in performance is obtained. Application of this principle and the corresponding performance and computational complexity tradeoffs are discussed for adaptive beamforming and detection, Volterra filtering, and estimation of higher order statistics.
Document Details
- Document Type
- Technical Report
- Publication Date
- Mar 12, 1997
- Accession Number
- ADA324997
Entities
People
- Barry Van Veen
Organizations
- University of Wisconsin–Madison