Divergence measures for time-frequency distributions

Distance measures between statistical models or between a model and observations are widely used concepts in signal processing. They are commonly used in solving problems such as detection, automatic segmentation, classification, pattern recognition and coding. In recent years, there has been an interest in extending these distance measures to the time-frequency plane. It has been suggested that these measures can be used for discriminating between nonstationary signals based on their time-frequency representations. In this paper, several well-known distance measures from information theory will be adapted to the time-frequency plane. The application of these measures for signal detection will be presented. The performance of these measures will be illustrated through an example.