Multi-Feature Multi-Sensor Fusion for Emitter Identification Based on a Modified DS Application

Abstract

Emitter identification is a crucial task in electronic countermeasure technology area, which deeply affects the accuracy of subsequent threat estimation. In emitter identification system, sensors (transmitter and receiver) have inevitable inconsistency and fuzziness, along with possible ambiguity and instability under interference and malfunction. To manage the uncertainty in emitter identification system, we propose a multi-feature multi-sensor fusion algorithm based on a modified DS application. The modified DS application for emitter identification system is accomplished by two parts—multi-feature fusion based on the improved proximity approach to obtain basic probability assignment ( $BPA$ ) and multi-sensor fusion based on the combination of two revised evidences to solve potential evidence conflicts. Firstly, the multi-feature fusion method based on the improved proximity approach is raised to produce the identification result for each receiver, which simultaneously build $BPAs$ for DS application. Four entropies are extracted to establish the multi-feature description of received signal. Then, in order to solve the inconsistency among different receivers and realize multi-sensor fusion for emitter identification, the multi-sensor fusion method based on the combination of two revised evidences is proposed. Two revised evidences are put forward, respectively, by the introduction of Lance distance function and spectral angle cosine function before applying DS combination. Experiments and analyses comprehensively demonstrate the great uncertainty management performance and favorable emitter identification effect of the proposed algorithm.