Representation of antenna calibration data using modular neural networks

To perform accurate AOA (angle of arrival) estimations using the MUSIC (MUltiple SIgnal Classification) algorithm, accurate antenna array calibration data must be available. Since it is not feasible to store calibration data for all possible AOAs, developing continuous functions to represent the cal...

Full description

Saved in:
Bibliographic Details
Main Authors: Niven, J., Teague, K.A.
Format: Conference Proceeding
Language:English
Subjects:
Calibration
Classification algorithms
Function approximation
Interpolation
Iterative algorithms
Linear antenna arrays
Multiple signal classification
Neural networks
Neurons
Transfer functions
Online Access:Request full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:To perform accurate AOA (angle of arrival) estimations using the MUSIC (MUltiple SIgnal Classification) algorithm, accurate antenna array calibration data must be available. Since it is not feasible to store calibration data for all possible AOAs, developing continuous functions to represent the calibration data is an attractive alternative. Neural networks are effective for estimating functions, and are very applicable for this situation. Two different iterative training methods for modular networks are presented and compared. The first method employs a fixed section size and variable network size, and the second method employs a variable section size and a fixed network size.
DOI:10.1109/MWSCAS.2002.1186984