Abstract
Classification of personnel targets by micro-Doppler signatures has received a growing interest in recent years. Although most of the work has been carried out in the RF regime for radar systems, much less research has been done in acoustic.
In this thesis a dataset collected with an acoustic radar, which consists of micro-Doppler signatures of different personnel targets undertaking various motions, is analysed. A set of heuristic features, such as the micro-Doppler signature bandwidth and period, is extracted from the data and it is used together with Cepstral features to assess classification performance of a Maximum Likelihood classifier.
Results show that high level classification performance can be achieved for both human recognition and motion classification.