TY - GEN
T1 - Detection and estimation of generalized chirps using time-frequency representations4
AU - Papandreou, Antonia
AU - Boudreaux-Bartels, G. Faye
AU - Kay, Steven M.
N1 - Funding Information:
'Funding by ONR grant N00014-92-5-1499,and by AFOSR 93-1-0006.
Publisher Copyright:
© 1995 IEEE.
PY - 1994
Y1 - 1994
N2 - We propose techniques for the detection and parameter estimation of generalized chirps in the presence of noise. Generalized chirps are nonstationary signals characterized by group delays with specific dispersion law characteristics. Special cases of generalized chirps include linear chirps, and hyperbolic chirps that are Doppler-invariant signals. We optimally detect generalized chirps using generalized timeshift covariant quadratic time-frequency representations (QTFRs) such as hyperbolic QTFRs used for detecting hyperbolic chirps. We also propose the parameter estimation of generalized chirps, and specialize our simulation results to hyperbolic chirps. We combine phase unwrapping with linear regression of the phase data at high signal-to-noise ratios (SNRs) to produce very simple and unbiased estimators that attain the Cramer-Rao lower bounds on variance. Maximum likelihood estimation performs well at low SNRs, but at the cost of high computational complexity.
AB - We propose techniques for the detection and parameter estimation of generalized chirps in the presence of noise. Generalized chirps are nonstationary signals characterized by group delays with specific dispersion law characteristics. Special cases of generalized chirps include linear chirps, and hyperbolic chirps that are Doppler-invariant signals. We optimally detect generalized chirps using generalized timeshift covariant quadratic time-frequency representations (QTFRs) such as hyperbolic QTFRs used for detecting hyperbolic chirps. We also propose the parameter estimation of generalized chirps, and specialize our simulation results to hyperbolic chirps. We combine phase unwrapping with linear regression of the phase data at high signal-to-noise ratios (SNRs) to produce very simple and unbiased estimators that attain the Cramer-Rao lower bounds on variance. Maximum likelihood estimation performs well at low SNRs, but at the cost of high computational complexity.
UR - http://www.scopus.com/inward/record.url?scp=35148827196&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=35148827196&partnerID=8YFLogxK
U2 - 10.1109/ACSSC.1994.471415
DO - 10.1109/ACSSC.1994.471415
M3 - Conference contribution
AN - SCOPUS:35148827196
T3 - Conference Record - Asilomar Conference on Signals, Systems and Computers
SP - 50
EP - 54
BT - Conference Record - 28th Asilomar Conference on Signals, Systems and Computers, ACSSC 1994
PB - IEEE Computer Society
T2 - 28th Asilomar Conference on Signals, Systems and Computers, ACSSC 1994
Y2 - 31 October 1994 through 2 November 1994
ER -