Algorithm 1: Multiaspect SVM

Input: Preprocessed radar training data

Output: Log-likelihood ratio (LLR) for each target

TRAINING:

for i:M aspects do

Perform training of SVM: wi, bi

Compute target probabilities, Pi(X|O) from PPV and NPV statistics

Carry out Laplace smoothing: ${\widehat{\theta }}_i$

end for

Sort aspects from highest value to lowest value

TESTING:

Determine aspects required for data collection: N

for i:N aspects do

for j:P testPoints do

Make radar measurement

Preprocess data: x

Perform classification: h(x)

Lookup Pj(X|O) from training data for each target

Apply sum rule to determine average Pi(X|O) for each target

end for

Apply product rule to determine aggregate: P(X|O) for each target

Update LLR

end for