Binary Hypothesis Testing: Example

H 0 : x[n]  w[n] n  0,1,..., N  1

H1 : x[n]  A  w[n] n  0,1,..., N  1
Receiver Operating Characteristics (ROC)
Receiver Operating Characteristics (ROC)
…….
Receiver Operating Characteristics (ROC)
…….
Receiver Operating Characteristics (ROC)
…….
Minimum Probability of Error
• In some detection problems on can reasonably assign probabilities to the various
hypotheses.
• In doing so, we express a belief in the likelihood of the hypotheses.
• This type of approach, where we assign prior probabilities, is the Bayesian approach to
the hypotheses testing.
• It is completely analogous to the Bayesian philosophy of estimation in which a prior
PDF is assigned to an unknown parameter
• With the Bayesian paradigm, we can define a probability of error Pe as
Pe = Pr{decide H0, H1 true} + Pr{decide H1, H0 true}
= P(H0 |H1)P(H1) + P(H1 |H0)P(H0)
• P(Hi ;Hj) is the probability of deciding Hi if Hj is true with no probabilistic meaning
assigned to the likelihood that Hj is true
• P(Hi |Hj) assumes that the outcome of a probabilistic experiment is observed to be Hj
and that the probability of deciding Hi is conditioned on the outcome.
• Using the Pe criterion, our goal is to design a detector that minimizes Pe.
Minimum Probability of Error ………
• In the derivation for the minimum Pe detector, we should decide H1 if

• If the prior probabilities are equal, we decide H1 if

p(x|H1) > p(x|H0)
i.e., we choose the hypothesis with the larger conditional likelihood or
the one that maximizes p(x|Hi) for i = 0,1. This is called the maximum
likelihood (ML) detector.
Bayesian Risk
Bayesian Risk …….
Detection of Deterministic Signals
Known Signal in Gaussian Noise – Matched Filter
SNR Maximization by the Matched Filter
Performance of Matched Filter
Generalized Matched Filter
Correlated Noise and Whitening
Performance of Generalized Matched Filter