TY - JOUR
T1 - Threshold setting for adaptive matched filter and adaptive coherence estimator
AU - Liu, Jun
AU - Li, Hongbin
AU - Himed, Braham
PY - 2015/1
Y1 - 2015/1
N2 - It is known that the probabilities of false alarm (PFAs) of several celebrated adaptive detectors including the adaptive matched filter (AMF) and the adaptive coherence estimator (ACE) can be expressed as integral forms. Nevertheless, it is inconvenient to set the detection thresholds by using these integral expressions. Here, we propose two computationally efficient schemes to calculate the thresholds of the AMF and ACE. In the first method, approximate expressions, in forms of elementary functions, for the PFAs of the AMF and ACE are derived. The thresholds of the AMF and ACE can be numerically computed by using these elementary expressions instead of the integrals, for reducing computational complexity. In the second approach, further approximations are employed to lead to highly simple expressions for the thresholds of the AMF and ACE, which enable us to directly compute the thresholds for a given PFA. Compared to the first one, the second scheme is more computationally efficient, but at the cost of a slight loss in accuracy. Numerical results verify the effectiveness of the two proposed schemes.
AB - It is known that the probabilities of false alarm (PFAs) of several celebrated adaptive detectors including the adaptive matched filter (AMF) and the adaptive coherence estimator (ACE) can be expressed as integral forms. Nevertheless, it is inconvenient to set the detection thresholds by using these integral expressions. Here, we propose two computationally efficient schemes to calculate the thresholds of the AMF and ACE. In the first method, approximate expressions, in forms of elementary functions, for the PFAs of the AMF and ACE are derived. The thresholds of the AMF and ACE can be numerically computed by using these elementary expressions instead of the integrals, for reducing computational complexity. In the second approach, further approximations are employed to lead to highly simple expressions for the thresholds of the AMF and ACE, which enable us to directly compute the thresholds for a given PFA. Compared to the first one, the second scheme is more computationally efficient, but at the cost of a slight loss in accuracy. Numerical results verify the effectiveness of the two proposed schemes.
KW - Adaptive coherence estimator
KW - Laplace approximation
KW - Rao test
KW - adaptive detection
KW - adaptive matched filter
KW - generalized likelihood ratio test
UR - http://www.scopus.com/inward/record.url?scp=84906245223&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=84906245223&partnerID=8YFLogxK
U2 - 10.1109/LSP.2014.2345757
DO - 10.1109/LSP.2014.2345757
M3 - Article
AN - SCOPUS:84906245223
SN - 1070-9908
VL - 22
SP - 11
EP - 15
JO - IEEE Signal Processing Letters
JF - IEEE Signal Processing Letters
IS - 1
M1 - 6875898
ER -