4
Implementation details
• Assumptions:
– M = 2 always detect 2 objests from antenna samples. Will also calculate confidence
metric for following module to filter results.
– Receive antennas are equally spaced
• Calculation of Π
A
(θ) can be significantly reduced with a(θ) constructed
for equally spaced antenna. It becomes an Hermitian matrix that is also
persymmetric.
• Given that R
n
is also Hermitian, calculation of {trace(Π
A
(θ)*R
n
) can also be
simplified.
• Total cycle cost depends on the length of steering vector a(θ).
– 2 stage search: first stage of coarse search, then zoom in with finer granularity in
second stage search.
• Output of normalized variance of per target (dimension) metrics
– The smaller the normalized variance, the stronger the signal from the
particular target
– Experimenting a threshold to determine single target.
