Abstract: Methods and systems, including computer programs encoded on a computer-readable medium, are described for implementing geolocation-aided unique signal recognition. For example, a system obtains radio frequency (RF) signals emitted by one or more emitters and processes the RF signals using a deinterleaving operation that integrates geolocation information for individual receivers that detect at least one of the RF signals. A grouping of RF signals that indicate an association with a particular one of the emitters is determined based on the deinterleaving operation. The system i) generates iterative sets of labeled training data with labels for RF signal inputs derived from the signal grouping and ii) generates an RF signal recognition model from machine-learning performed using the iterative sets of labeled training data. Based on detection of an RF signal by a receiver, the RF recognition model determines a geolocation of a corresponding emitter that emits a particular RF signal.

Abstract: Systems and methods to implement the initialization and steady state operation of a system for geolocation-aided unique signal recognition (USR). The unique signal recognition (USR) technique uses geospatial location data (geolocation data), labeled datasets, expert curated datasets, and partially labeled “enhanced” datasets for training a machine-learning radio frequency signal recognition model (also referred to as “RF recognition model” (RFRM)) to recognize and locate certain RF signal emitters of interest. The unique signal recognition may utilize a variety of data clustering techniques that operate on batches of data, comprising a corpus of linked RF data and derived features.

Abstract: A method includes training a machine learning network using, as input training data, a plurality of time series representing sampled radio frequency (RF) signals, and a plurality of labels. Each label is associated with at least one time series of the plurality of time series. The labels describe one or more characteristics of emitters of the sampled RF signals. The machine learning network includes, for each label of the plurality of labels, a corresponding soft dynamic time warping (soft-DTW) terminal associated with the label. Training the machine learning network includes iteratively adjusting the weights of the soft-DTW terminals so as to reduce a value of at least one loss function; obtaining, as a result of the training, a machine learning model configured to classify new RF signal sequences by label; and providing the machine learning model to an RF sensing device.

Abstract: Systems and methods to implement the initialization and steady state operation of a system for geolocation-aided unique signal recognition (USR). The unique signal recognition (USR) technique uses geospatial location data (geolocation data), labeled datasets, expert curated datasets, and partially labeled “enhanced” datasets for training a machine-learning radio frequency signal recognition model (also referred to as “RF recognition model” (RFRM)) to recognize and locate certain RF signal emitters of interest. The unique signal recognition may utilize a variety of data clustering techniques that operate on batches of data, comprising a corpus of linked RF data and derived features.