Information Theoretic Measures for Through-the-Wall Surveillance

Two information theoretic measures, entropy and divergence, are considered for detecting possible scene variations in the emerging through-the-wall radar imaging and surveillance applications. Investigation of these measures shows that they are likely candidates to provide automated reliable notifications of single and multiple changes in the scene. This capability is a key and welcome feature for practical and effective through-the-wall surveillance systems. Using experimental data, we show that the entropy measure is useful for discriminating between populated and unpopulated settings, whereas the divergence measure should be applied by the system operator for monitoring changes in the scene