The proverbial multi-INT firehose is well documented. There is simply too much data for any team of human analysts to effectively digest and process in real (actionable) time. Consequently, for many years now, there have been numerous attempts to automate multi-sensor exploitation techniques so as to reduce operator(s) workload.Recent advances in machine intelligence and deep learning have rekindled interest in developing automated multi-INT fusion based on these new emerging techniques. Of course, if machine intelligence techniques are developed and applied, a complete re-work of the multi-INT architecture is warranted to jointly optimize collection and automated deep learning exploitation.Recent successes of deep learning neural network techniques and architectures have been well publicized over the last several years. They exploit correlations (sometimes subtle) that lead to successful decisioning. Patterns in the input are processed to reveal correlations that were successful during the training process.We bring together advances in multiphysics-based sensor fusionand deep learning techniques to provide an entirely new approach to both the design and operation of distributed multi-INT sensor systems. This approach is in contrast to conventional multi-INT fusion engines that fuse post-measurement sensor products such as features from each stovepiped sensor.Deep Learning Neural Networks,convolutional neural networks,Recurrent Neural Networks,long short term memory,Multi-INT data fusion,SIGINT,multi-physics,sensor fusion
