Page 130 - Radar Technology Encyclopedia

120 detection criterion, minimax detection, noncoherent

lates the average risk r(P ) for the worst values of a priori tion cell, and employing a binary detector. If the output volt-
aj
probabilities P , j = 0, 1, ... , k, for which age of the Kth channel exceeds the threshold, then the output
aj
r(P ) ³ r(P ), j = 0, 1, ... , K, voltage from the threshold unit 2 goes to decision-making
a
aj
where r(P ) is the average risk based on the minimum aver- unit 3, where the decision about the target presence is made
a
age risk criterion. AIL and the proper cell where detection took place is determined.
Ref.:DiFranco (1968), p. 274; Lukoshkin (1983), p. 28. Typically, multichannel detection is used in phased-array
radars. AIL
Missed detection refers to the erroneous decision that a sig-
nal is absent when it is actually present. In statistical detection Ref.: Lukoshkin (1983), pp. 38–43.
theory it is called a type II error (as opposed to type I error Input
that is a false alarm). SAL
Ref.: Skolnik (1980), p. 376.
Signal Signal Signal
processing processing processing
moving-target detector (see MOVING-TARGET DETEC- channel 1 channel 1 channel 1
TOR).
Moving-window detection examines continuously the last n Threshold Threshold Threshold
unit 2 unit 2 unit 2
samples within each quantized range interval and announces
the presence of a target if m-out-of-n samples cross a preset
threshold. The block diagram of binary moving-window Decision-making unit 3
detection is shown in Fig. D23. Such a procedure is less effi-
Output
cient than ideal postdetection integration, but it has the advan-
Figure D24 Multichannel detector.
tage of simplicity.
In binary moving-window detection, the test statistic S is Multiple-pulse detection uses the pulse train received from
j
computed using the target. The technique that improves detection perfor-
mance in this case is integration. A typical configuration for
j
this type of detector consists of a detector matched to an indi-
S = å y
j k
vidual pulse in the train, followed by a coherent or noncoher-
–
k = j M + 1 ent integrator, depending on the structure of the transmitted
where y is the received sample and M is the window length. waveform and the radar (see also INTEGRATOR). SAL
k
To compute and threshold the statistics S , the binary moving- Ref.: Blake (1980), p. 41.
j
window detection can use, for example, a digital delay line,
counter, and comparator. The detection decision is made The Neyman-Pearson (detection) criterion is the optimum
when S crosses a threshold K in the upward direction: in detection criterion giving the maximum probability of detec-
m
j
effect, when S j- 1 < K and S = K . This detection process is tion P value. It requires that the acceptable probability of
d
m
m
j
often termed m-out-of-n detection, sliding-window detection, false alarm P be assigned. Then the best is considered that
fa
or binary integration. SAL detector giving the maximum P under the condition that the
d
P will exceed the specified value
Ref.: Skolnik (1980), p. 388; Neri (1991), p. 121. fa
P = max if P £ P fa0
fa
d
AIL
Ref.: Druzhinin (1967), p. 532; Barkat (1991), pp. 145–147.
Noncoherent detection is the detection of a noncoherent sig-
nal in an interference background. For the noncoherent pulse
train, due to the statistical independence of the initial phases
of the pulses and the fact that the noise is not correlated from
pulse to pulse, the likelihood ratio of the train L is equal to
the product of likelihood ratios for each pulse. If we take the
logarithm of L:
Figure D23 Block diagram of a binary moving-window detec- n
tion (after Skolnik, 1980, Fig. 10.7, p. 388). l = ln L = ln ( L × L × å ln L
¼ L× ) =
1 2 n i
multinomial [multiple-coincidence] detection (see coinci- i = 1
dence detection). the algorithm for optimum noncoherent detection can be cited
as
Multichannel detection is used in a multichannel radar. The
l ³ then target is present.
z
block diagram of a multichannel detector when surveillance 0
l < z then target is absent.
is performed with a single beam is given on Fig. D24. There 0
where z is the threshold defined by the selected detection cri-
are K detection channels, each corresponding to one resolu- 0
terion.