System and method for time-frequency separation of multiple radio signals

Abstract

1. A system for time-frequency separation based on image analysis of one or more spectrograms of a plurality of wireless signals received over time from one or more emitters, the system comprising: a processing circuit configured to: determine, using the image analysis, a region of at least one of the spectrograms that contains a representation of a superposition of two or more overlapping wireless signals of the wireless signals, the overlapping wireless signals being representations of wireless signals that share a common time frequency in at least one of the spectrograms; decompose the superposition of the two or more overlapping wireless signals in the region into separate signal representations, each given one of the separate signal representations being represented by a corresponding bounding box around the given separate signal representation; and perform one or more actions using the separate signal representations.

Description

[Technical Field]

[0001] The present invention relates to a system and method for time-frequency separation of multiple radio signals received over time from one or more emitters. [Background technology]

[0002] Wideband RF systems that receive and analyze radio frequency (RF) communications at multiple frequencies are important in many domains, including automotive radar solutions, cellular and Wi-Fi coverage analysis, and communications intelligence systems. Current wideband RF analysis utilizes one-dimensional signal processing methods to analyze the energy levels of intercepted RF communications emitted at multiple frequencies from multiple sources to detect areas within the intercepted RF communications that contain signals of interest to users of the wideband RF system. Current wideband RF systems are designed to process communications emitted from a limited number of narrowband, static or cyclically varying frequencies, and on a standard grid with known modulations (e.g., amplitude modulation (AM), frequency modulation (FM), frequency shift keying (FSK), etc.). Current wideband RF systems are not well suited to analyzing complex RF communications emitted from multiple sources, which may be wideband and employ adaptive frequency-hopping spread-spectrum communication techniques, such as those used in many modern RF transceivers in ad hoc networks. These current wideband RF analysis systems are not suited to handling bursty communication sources, wideband messages (such as video feeds and multimedia feeds), multiple communication sources, sources with advanced modulation, and rich communication protocols.

[0003] In particular, current wideband RF systems receive multiple signals from multiple communication sources. These incoming signals can interfere with each other, causing the wideband RF system to receive superimposed signals. Current wideband RF analysis systems that utilize one-dimensional signal processing methods to separate the superimposed signals into multiple incoming signals have limited performance and separation capabilities.

[0004] Therefore, there is a need in the art for new methods and systems for wideband RF analysis, and in particular for time-frequency separation of one or more spectrograms of multiple wireless signals received over time from one or more emitters that may be unknown, wideband, use burst communications, use complex modulation, or have no standard / pre-identified grid.

[0005] The following references are believed to be relevant as background to the presently disclosed subject matter. The acknowledgment of a reference herein should not be inferred as meaning that it is in any way relevant to the patentability of the presently disclosed subject matter.

[0006] Patent Document 1 discloses a signal identification method based on a wireless signal spectral feature template. The method includes: constructing a signal template library by extracting spectral features of multiple wireless signals; preprocessing the signal to be identified according to the representation format of the signal template in the signal template library; and comparing and matching the preprocessed signal to be identified with the signal spectrum template library to obtain the type and spectral feature parameters of the signal to be identified. The present invention also provides a signal identification system based on a wireless signal spectral feature template, comprising a template construction unit, a preprocessing unit, and a matching unit. In the present invention, the spectral templates can represent various types of signals, and the signal identification algorithm is further simplified while ensuring the accuracy of signal identification.

[0007] Non-Patent Document 1 discloses that modulation identification is a key function for intelligent receivers. Numerous applications exist in cognitive radar, software-defined radio (SDR), and efficient spectrum management. To identify communication and radar waveforms, their modulation types must be classified. DARPA's Spectrum Coordination Challenge highlights the need to manage demand for shared RF spectrum. Here, we show how learning techniques can be leveraged in these types of applications to effectively identify modulation schemes.

[0008] Non-Patent Document 2 discloses a novel framework based on image processing techniques for radio environment characterization. In particular, after digitizing a given scenario according to the frequencies detected at each point, an image can be constructed in which the pixel intensities of the image capture the radio-electrical conditions. Through non-linear filtering and object detection operations, the proposed framework makes it possible to identify homogeneous regions in which several frequencies can be detected. This method can be used in various contexts, such as building a database to provide spectrum awareness to terminals in flexible spectrum scenarios to facilitate switch-on procedures or secondary spectrum utilization, and supporting the devising of appropriate strategies for cell and RAT selection.

[0009] Non-Patent Document 3 discloses that cognitive radio is a new paradigm for wireless communications that offers a solution for reconciling current spectrum underutilization with growing demand without modifications to existing legacy wireless systems. Secondary users should be able to identify spatial and temporal spectrum holes not occupied by primary users and opportunistically use them without causing interference to primary receivers. To this end, knowledge of the primary network is required to ensure proper secondary user operation. In this context, assuming no collaboration between the primary and secondary networks, this specification proposes a new framework based on image processing techniques that aims to combine several sensed samples at different geographic locations collected by secondary sensors to estimate the locations of different primary transmitters. The results can be used to discover frequencies that the secondary network can use without interfering with the primary receiver. Our results, performed in a realistic scenario, demonstrate the effectiveness of the proposed framework in estimating transmitter locations.

[0010] Patent Document 2 discloses a flutter signal analysis method based on a convolutional neural network and a short-time Fourier transform, which includes the steps of performing time-frequency analysis on an actually measured flutter signal by using a short-time Fourier transform to obtain a time-frequency graph of the flutter signal, then mining image characteristics by using the powerful image processing capabilities of a convolutional neural network, extracting flutter characteristics, and subsequently analyzing the signal by calculating through sufficient connection layers and a loss function. This invention combines a convolutional neural network with a short-time Fourier transform of a flutter signal, which has good reliability and accuracy for analyzing actually measured flutter data, laying a solid foundation for further development of research on combining artificial intelligence with aeroelasticity, and has practical engineering application value. [Prior art documents] [Patent documents]

[0011] [Patent Document 1] International Publication No. 2016 / 082562 [Patent Document 2] Chinese Patent Application Publication No. 110866448 [Non-patent literature]

[0012] [Non-Patent Document 1] Algorithms to antenna: train deep-learning networks with synthesized radar and comms signals (Gentile et al.) published on November 20, 2019 [Non-patent document 2] On the applicability of image processing techniques in the radio environment characterization (Perez-Romero et al.) published on April 2009 [Non-patent document 3] Image processing techniques as a support to transmitter positioning determination in cognitive radio networks (Bolea et al.) published on May 2010 Summary of the Invention [Means for solving the problem]

[0013] According to a first aspect of the presently disclosed subject matter, there is provided a system for time-frequency separation based on image analysis of one or more spectrograms of a plurality of radio signals received over time from one or more emitters, the system comprising: processing circuitry configured to: use the image analysis to determine a region of at least one of the spectrograms containing a representation of a superposition of two or more overlapping radio signals of the radio signals, where the overlapping radio signals are representations of radio signals that share a common time frequency in at least one of the spectrograms; decompose the superposition of the two or more overlapping radio signals in the region into separate signal representations, where each given one of the separate signal representations is represented by a corresponding bounding box around the given separate signal representation; and perform one or more actions using the separate signal representations.

[0014] In some cases, the processing circuitry is further configured to determine, for at least one of the distinct signal representations, one or more parameters based on a corresponding bounding box, and the one or more actions are also performed based on the parameters.

[0015] In some cases, the parameters of the distinct signal representation include one or more of: (a) wavelength, (b) frequency, (c) rise time, (d) fall time, (e) duration, (f) power, (g) occupied bandwidth, (h) spectral density, (i) activity profile, or (j) polarization.

[0016] In some cases, the processing circuitry is further configured to: generate a plurality of additional spectrograms of the plurality of wireless signals when decomposing the superposition of two or more overlapping wireless signals, wherein (a) a first additional spectrogram of the additional spectrograms has a first resolution, (b) a second additional spectrogram of the additional spectrograms has a second resolution, and (c) the first resolution is different from the second resolution; and, for at least one given distinct signal representation of the distinct signal representations, determine one or more fine-tuned parameters of the given distinct signal representation having a higher accuracy than the accuracy of the corresponding parameters using the additional spectrogram and the corresponding parameters.

[0017] In some cases, the image analysis of spectrograms of multiple radio signals received over time is performed continuously or periodically.

[0018] In some cases, the processing circuitry is further configured, upon decomposing the superposition of two or more overlapping radio signals, to determine one or more statistical parameters associated with at least one given distinct signal representation of the distinct signal representations based on prior occurrences of the given distinct signal representation in the spectrogram.

[0019] In some cases, the statistical parameters include one or more of (a) signal activity level, (b) probability of occurrence, (c) probability of interception, (d) power distribution, (e) frequency stability, or (f) average duration.

[0020] In some cases, the actions include one or more of: (a) identifying emitters that emitted overlapping wireless signals; (b) classifying emitters that emitted overlapping wireless signals; (c) generating a condensed representation of multiple wireless signals received over time; (d) extracting samples of the overlapping wireless signals, thereby enabling wireless signal analysis; (e) providing the overlapping wireless signals to one or more external systems; or (f) providing the overlapping wireless signals to a user of the system.

[0021] In some cases, the decomposition of the superposition of overlapping radio signals is based on Rectilinear Polygon Decomposition (RPD).

[0022] In some cases, the decomposition of the superposition of overlapping radio signals is based on artificial neural networks (ANNs).

[0023] In some cases, the determination of the region is also based on bands of multiple radio signals received over time.

[0024] In some cases, the region is determined by identifying a bounding box around the region.

[0025] In some cases, each of the distinct signal representations is represented by a corresponding bounding box around the distinct signal representation.

[0026] In some cases, spectrograms are used to optimize the signal-to-noise ratio (SNR) of multiple wireless signals.

[0027] According to a second aspect of the presently disclosed subject matter, there is provided a method for time-frequency separation based on image analysis of one or more spectrograms of multiple radio signals received over time from one or more emitters, the method including: determining, by a processing circuit, using the image analysis, a region of at least one of the spectrograms containing a representation of a superposition of two or more overlapping radio signals of the radio signals, where the overlapping radio signals are representations of radio signals that share a common time frequency within at least one of the spectrograms; decomposing, by the processing circuitry, the superposition of the two or more overlapping radio signals in the region into separate signal representations, where each given one of the separate signal representations is represented by a corresponding bounding box around the given separate signal representation; and performing, by the processing circuitry, one or more actions using the separate signal representations.

[0028] In some cases, the method further includes determining, by the processing circuitry, one or more parameters for at least one of the distinct signal representations based on a corresponding bounding box, wherein one or more actions are also performed based on the parameters.

[0029] In some cases, the parameters of the distinct signal representation include one or more of: (a) wavelength, (b) frequency, (c) rise time, (d) fall time, (e) duration, (f) power, (g) occupied bandwidth, (h) spectral density, (i) activity profile, or (j) polarization.

[0030] 16. The method of claim 15, further comprising: optionally generating, by the processing circuitry, a plurality of additional spectrograms of the plurality of wireless signals upon decomposing the superposition of two or more overlapping wireless signals, wherein (a) a first additional spectrogram of the additional spectrograms has a first resolution, (b) a second additional spectrogram of the additional spectrograms has a second resolution, and (c) the first resolution is different from the second resolution; and, for at least one given distinct signal representation of the distinct signal representations, determining, by the processing circuitry, one or more fine-tuned parameters of the given distinct signal representation using the additional spectrogram and corresponding parameters, the fine-tuned parameters having a precision greater than a precision of the corresponding parameters.

[0031] In some cases, the image analysis of spectrograms of multiple radio signals received over time is performed continuously or periodically.

[0032] In some cases, upon decomposing the superposition of two or more overlapping wireless signals, the method further includes determining one or more statistical parameters associated with at least one given distinct signal representation of the distinct signal representations based on prior occurrences of the given distinct signal representation in the spectrogram.

[0033] In some cases, the statistical parameters include one or more of (a) signal activity level, (b) probability of occurrence, (c) probability of interception, (d) power distribution, (e) frequency stability, or (f) average duration.

[0034] In some cases, the actions include one or more of: (a) identifying emitters that emitted overlapping wireless signals; (b) classifying emitters that emitted overlapping wireless signals; (c) generating a condensed representation of multiple wireless signals received over time; (d) extracting samples of the overlapping wireless signals, thereby enabling wireless signal analysis; (e) providing the overlapping wireless signals to one or more external systems; or (f) providing the overlapping wireless signals to a user of the system.

[0035] In some cases, the decomposition of the superposition of overlapping radio signals is based on rectilinear polygonal decomposition (RPD).

[0036] In some cases, the decomposition of the superposition of overlapping radio signals is based on artificial neural networks (ANN).

[0037] In some cases, the determination of the region is also based on bands of multiple radio signals received over time.

[0038] In some cases, the region is determined by identifying a bounding box around the region.

[0039] In some cases, each of the distinct signal representations is represented by a corresponding bounding box around the distinct signal representation.

[0040] In some cases, the spectrogram is used to optimize the signal-to-noise ratio (SNR) of multiple wireless signals.

[0041] According to a third aspect of the presently disclosed subject matter, there is provided a non-transitory computer-readable storage medium having computer-readable program code embodied therein, the computer-readable program code being executable by at least one processing circuit of a computer to perform a method for time-frequency separation based on image analysis of one or more spectrograms of a plurality of wireless signals received over time from one or more emitters, the method including: determining, by the processing circuit, using the image analysis, a region of at least one of the spectrograms containing a representation of a superposition of two or more overlapping wireless signals of the wireless signals, wherein the overlapping wireless signals are representations of wireless signals that share a common time frequency in at least one of the spectrograms; decomposing, by the processing circuit, the superposition of the two or more overlapping wireless signals in the region into separate signal representations, wherein each given one of the separate signal representations is represented by a corresponding bounding box around the given separate signal representation; and performing, by the processing circuit, one or more actions using the separate signal representations.

[0042] In order to understand the presently disclosed subject matter and to see how it may be carried out in practice, the subject matter will now be described, by way of non-limiting example only, with reference to the accompanying drawings, in which: [Brief explanation of the drawings]

[0043] [Figure 1A] 1 is a schematic diagram of an exemplary input spectrogram having a representation of one or more input signals in accordance with the presently disclosed subject matter. [Figure 1B] 1 is a schematic diagram of an example spectrogram with a representation of one or more of the input signals above a threshold in accordance with the presently disclosed subject matter. [Figure 1C] 1 is a schematic diagram of an example spectrogram with a representation of one or more identified groups of pixels in accordance with the presently disclosed subject matter. [Figure 1D] 1 is a schematic diagram of an example spectrogram with a representation of one or more groups with marked boundaries in accordance with the presently disclosed subject matter; [Figure 2] FIG. 1 is a block diagram that schematically illustrates one example of a system for time-frequency separation based on image analysis of one or more spectrograms of multiple radio signals received over time from one or more emitters, in accordance with the presently disclosed subject matter. [Figure 3] 1 is a flowchart illustrating one example sequence of operations performed for time-frequency separation based on image analysis of one or more spectrograms of multiple radio signals received over time from one or more emitters in accordance with the presently disclosed subject matter. [Figure 4A] 1 is a schematic diagram of an example region within a spectrogram having representations of one or more overlapping signals in accordance with the presently disclosed subject matter; [Figure 4B] 1 is a schematic diagram of an example region within a spectrogram having representations of one or more overlapping signals in accordance with the presently disclosed subject matter; [Figure 4C] 1 is a schematic diagram of an example region within a spectrogram having representations of one or more overlapping signals in accordance with the presently disclosed subject matter; [Figure 4D] 1 is a schematic diagram of an example region within a spectrogram having representations of one or more overlapping signals in accordance with the presently disclosed subject matter; [Figure 4E] 1 is a schematic diagram of an example region within a spectrogram having representations of one or more overlapping signals in accordance with the presently disclosed subject matter; [Figure 4F] 1 is a schematic diagram of an example region within a spectrogram having representations of one or more overlapping signals in accordance with the presently disclosed subject matter; [Figure 4G] 1 is a schematic diagram of an example region within a spectrogram having representations of one or more overlapping signals in accordance with the presently disclosed subject matter; [Figure 4H] 1 is a schematic diagram of an example region within a spectrogram having representations of one or more overlapping signals in accordance with the presently disclosed subject matter; [Figure 5] 1 is a schematic diagram of an example spectrogram having a representation of one or more separated signals in accordance with the presently disclosed subject matter. [Figure 6] 1 is a schematic diagram of an exemplary spectrogram with good time resolution having a representation of two exemplary separated signals, and an exemplary spectrogram with good frequency resolution having a representation of the two exemplary separated signals, in accordance with the presently disclosed subject matter. DETAILED DESCRIPTION OF THE INVENTION

[0044] In the following detailed description, numerous specific details are set forth in order to provide a thorough understanding of the presently disclosed subject matter. However, it will be understood by those skilled in the art that the presently disclosed subject matter may be practiced without these specific details. In other instances, well-known methods, procedures, and components have not been described in detail so as not to obscure the presently disclosed subject matter.

[0045] In the drawings and descriptions set forth, like reference numerals designate components that are common to different embodiments or configurations.

[0046] Unless otherwise specified, as will be apparent from the following description, it will be appreciated that throughout the description herein, the use of terms such as "determining," "resolving," "performing," "updating," "separating," and the like includes computational acts and / or processes that manipulate and / or transform data into other data, where said data are represented as physical quantities, e.g., electronic quantities, and / or where said data represent physical objects. The terms "computer," "processor," "processing resource," "processing circuitry," and "controller" should be interpreted broadly to cover any type of electronic device with data processing capability, including, by way of non-limiting example, personal desktop computers / personal laptop computers, servers, computing systems, communications devices, smartphones, tablet computers, smart televisions, processors (e.g., digital signal processors (DSPs), microcontrollers, field programmable gate arrays (FPGAs), application specific integrated circuits (ASICs), etc.), groups of multiple physical machines that share performance for various tasks, virtual servers co-resident on a single physical machine, other electronic computing devices, and / or any combination thereof.

[0047] Operations according to the teachings herein may be performed by a computer specially constructed for the desired purpose, or by a general-purpose computer specially configured for the desired purpose by a computer program stored on a non-transitory computer-readable storage medium. The term "non-transitory" is used herein to exclude transitory propagating signals, but is otherwise used to include volatile or non-volatile computer memory technology suitable for the application.

[0048] As used herein, the phrases "for example," "such as," "for example," and variations thereof refer to non-limiting examples of the presently disclosed subject matter. Reference herein to "in one instance," "in some instances," "in other instances," or variations thereof means that a particular feature, structure, or characteristic described in connection with an example is included in at least one example of the presently disclosed subject matter. Thus, appearances of the phrases "in one instance," "in some instances," "in other instances," or variations thereof do not necessarily refer to the same example.

[0049] It will be appreciated that, unless otherwise stated, certain features of the presently disclosed subject matter that are, for clarity, described in the context of separate embodiments, can also be provided in combination in a single embodiment. Conversely, various features of the presently disclosed subject matter that are, for brevity, described in the context of a single embodiment, can also be provided separately or in any suitable subcombination.

[0050] In embodiments of the presently disclosed subject matter, fewer, more, and / or different steps than those shown in FIG. 3 may be performed. In embodiments of the presently disclosed subject matter, one or more steps shown in FIG. 3 may be performed in a different order and / or one or more groups of steps may be performed simultaneously. FIGS. 1A-1D, 2, 4A-4H, and 5 show general schematic diagrams of system components and flow according to one embodiment of the presently disclosed subject matter. Each module and result shown in FIGS. 1A-1D, 2, 4A-4H, and 5 may be comprised of any combination of software, hardware, and / or firmware that performs the functions defined and described herein. The modules and results shown in FIGS. 1A-1D, 2, 4A-4H, and 5 may be centralized in one location or distributed across two or more locations. In other embodiments of the presently disclosed subject matter, the system may include fewer, more, and / or different modules than those shown in FIGS. 1A-1D, 2, 4A-4H, and 5.

[0051] References herein to a method should also apply mutatis mutandis to a system capable of carrying out the method, and to a non-transitory computer-readable medium storing instructions that, when executed by a computer, result in the method being performed.

[0052] References herein to a system should also apply mutatis mutandis to methods that may be performed by that system, and to non-transitory computer-readable media having instructions stored thereon that may be executed by that system.

[0053] References herein to non-transitory computer-readable media should be applied mutatis mutandis to a system capable of executing instructions stored on the non-transitory computer-readable media, and should also be applied mutatis mutandis to a method that can be executed by a computer reading instructions stored on the non-transitory computer-readable media.

[0054] With this in mind, attention is now directed to FIG. 1A, which is a schematic illustration of an exemplary input spectrogram having a representation of one or more input signals in accordance with the presently disclosed subject matter.

[0055] An input spectrogram is a two-dimensional image representing one or more input signals 110 of an intercepted communication. The input spectrogram may be collected by a wideband receiver (e.g., a spectrum analyzer) that intercepts various RF communications (e.g., wireless communications, cellular communications, Wi-Fi communications, radar communications, etc.) emitted by one or more emitters over a period of time. The input spectrogram may be analyzed using image analysis methods to detect areas within the intercepted communications that contain signals of interest for the wideband RF analysis system.

[0056] An input spectrogram is a graph in which the X-axis represents the time of interception (e.g., in milliseconds) and the Y-axis represents the frequency of interception (e.g., in MegaHertz (MHz)). The color (or gray scale) of each pixel located at a given X, Y location in the graph relates to the energy level (e.g., in decibels (dB)) intercepted by the wideband RF receiver at time X and frequency Y. A sequence of colored pixels may represent an input signal 110. For example, FIG. 1A shows a recurring sequence of light-colored pixel groups located on the Y-axis of the graph between pixel 1500 and pixel 2000. These pixel groups represent recurring input signal 110 with a given energy level (represented by the brightness level of these pixels) intercepted between pixel 1500 and pixel 2000.

[0057] The wideband RF analysis system can optionally process the input spectrogram to generate a filtered intercepted communications image, a non-limiting example of which is provided in FIG. 1B. A pixel of the filtered intercepted communications image located at a given X, Y location in the input spectrogram is in an "on state" if the color of the corresponding given pixel located at that X, Y location in the input spectrogram is above a color threshold (e.g., above a color representing -40 dBm). If the corresponding given pixel of the input spectrogram is below the color threshold, the pixel is "off state" in the filtered intercepted communications image and does not become part of further processing steps, as described in more detail herein. The pixels remaining in an "on state" represent the above-threshold signal 120. In the non-limiting example of FIG. 1B, input signals of the input signal 110 located at pixel 1500 at the start of the X-axis that are below the color threshold are removed and do not become part of the above-threshold signal 120.

[0058] Similarly, the wideband RF analysis system can optionally process the filtered intercepted communications image to generate a noise-filtered intercepted communications image, eliminating pixels above a noise threshold. The noise threshold can be determined for each spectrogram according to a desired sensitivity target. The filtering stage allows the wideband RF analysis system to process only the portion of the input spectrogram that has signals 120 above the threshold signal and is the noise-filtered signal, thereby providing better results for identifying signals within the input spectrogram.

[0059] The wideband RF analysis system processes the signal communication image above a threshold using a local distance grouping algorithm (e.g., a labeling algorithm that labels connected components in a 2D binary image) to detect groups of pixels within the filtered intercepted communication image. The wideband RF analysis system can also optionally filter out pixels not found by the local distance grouping algorithm that should be part of any group of pixels. Each identified group of pixels 130 can represent an identified signal within the intercepted communication, as described in further detail herein, particularly with reference to Figures 4A-4H, and the system can perform one or more actions on such signals, such as identifying overlapping wireless signals represented by groups of pixels that embody two or more overlapping wireless signals, represented as two or more subgroups of the identified group. A non-limiting example of the resulting processed image is shown in Figure 1C.

[0060] 1D is a non-limiting example of marking the identified group 130 with a boundary, thereby individually marking each of the groups of pixels, resulting in a group 140 with a marked boundary, which is a pixel identified as a group with a boundary around the identified group 130. Note that this phase of processing can optionally use a filtered intercepted communications image or the input spectrogram itself. Additionally, the described system can determine one or more parameters of the identified signal, such as wavelength, frequency, polarization, etc.

[0061] Having briefly described exemplary input spectrograms and the processing of the input spectrograms to identify one or more identified signals 140, attention is now directed to FIG. 2, which is a block diagram that schematically illustrates one example of a system for time-frequency separation based on image analysis of one or more spectrograms of multiple wireless signals received over time from one or more emitters, in accordance with the presently disclosed subject matter.

[0062] System 200 may comprise or otherwise be associated with a data repository 210 (e.g., a database, a storage system, a memory including read-only memory ROM, random access memory RAM, or any other type of memory, etc.) configured to store data including, among other things, input spectrograms, filtered intercepted communication images, noise-filtered intercepted communication images, processed images, thresholds, characteristics of identified signals 140, etc. In some cases, data repository 210 may be further configured to allow retrieval and / or update and / or deletion of data stored thereon. Note that in some cases, data repository 210 may be distributed. Note that in some cases, data repository 210 may be stored in cloud-based storage.

[0063] System 200 further includes a network interface 220, which allows system 200 to connect to a network and to send and receive data over the network, possibly including receiving one or more spectrograms representing RF signals collected over time, for example, from a wideband RF receiver. In a non-limiting sample setup, system 200 can analyze spectrograms in real time representing a 40 MHz spectrum containing multiple emitters emitting hundreds of signals per second, some of which may be burst communication emitters and / or frequency hopping emitters. In some cases, network interface 220 can connect to a local area network (LAN), a wide area network (WAN), or the Internet. In some cases, network interface 220 can connect to a wireless network.

[0064] System 200 further comprises processing circuitry 230. Processing circuitry 230 may be one or more processing circuitry units (e.g., central processing units), microprocessors, microcontrollers (e.g., microcontroller units (MCUs)), or any other computing device or module, including multiple and / or parallel and / or distributed processing circuitry units, adapted to process data, either singly or cooperatively, to control associated system 200 resources and to enable operations related to system 200 resources.

[0065] The processing circuit 230 comprises the following modules: a time-frequency separation module 240 .

[0066] The time-frequency separation module 240 may be configured to perform a time-frequency separation process as described in further detail herein, with particular reference to FIG.

[0067] FIG. 3 is a flowchart illustrating an example of a sequence of operations performed for time-frequency separation based on image analysis of one or more spectrograms of multiple radio signals received over time from one or more emitters in accordance with the presently disclosed subject matter.

[0068] According to some examples of the presently disclosed subject matter, system 200 can be configured to perform time-frequency separation process 300, utilizing, for example, time-frequency separation module 240, to identify and resolve overlapping radio signals represented in an input spectrogram. Overlapping radio signals can be collected by a wideband RF receiver when two or more emitters simultaneously emit radio signals within overlapping frequencies. The overlapping radio signals are represented by regions of the input spectrogram as groups of pixels that embody two or more overlapping radio signals. FIG. 4A is a schematic diagram of an example region (e.g., one of region A 410-a, region B 410-b, or region C 410-c) within the input spectrogram representing one or more overlapping signals according to the presently disclosed subject matter. System 200 analyzes at least one of the regions to determine whether the region contains superimposed representations of overlapping radio signals and resolves the overlapping signals into separate signal representations. To this end, system 200 can be configured to use image analysis to determine at least one region of a spectrogram (e.g., one of region A 410-a, region B 410-b, or region C 410-c) containing a representation of a superposition of two or more overlapping wireless signals, where the overlapping wireless signals are representations of wireless signals that share a common time frequency within the at least one of the spectrograms (block 310). System 200 analyzes the input spectrogram to determine at least one region within the input spectrogram (e.g., one of region A 410-a, region B 410-b, or region C 410-c). Determining the region can be achieved by using a machine learning model (e.g., an artificial neural network (ANN), random forest, decision tree, etc.) that masks all pixels associated with objects representing non-overlapping wireless signals, or by any other image analysis technique.The determination of the region (e.g., one of region A 410-a, region B 410-b, or region C 410-c) may also be based on bands (e.g., high frequency (HF) band, very high frequency (VHF) band, ultra high frequency (UHF) band, etc.) of multiple radio signals received over time and represented in the input spectrogram.

[0069] Note that a region (e.g., one of region A 410-a, region B 410-b, or region C 410-c) may be one or more groups of pixels from one or more areas in the input spectrogram, and these areas may be discontinuous areas of the input spectrogram. For example, one region (e.g., one of region A 410-a, region B 410-b, or region C 410-c) may include a group of pixels from the upper left corner of the input spectrogram and a group of pixels from the lower right corner of the input spectrogram.

[0070] After determining the region (e.g., one or more of region A 410-a, region B 410-b, or region C 410-c), system 200 may be further configured to decompose a superposition of two or more overlapping wireless signals within the region (e.g., one or more of region A 410-a, region B 410-b, or region C 410-c) into separate signal representations, each given one of the separate signal representations being represented by a corresponding bounding box around the given separate signal representation (block 320). The system 200 decomposes the superposition of two or more overlapping wireless signals in a region (e.g., one or more of region A 410-a, region B 410-b, or region C 410-c) by employing one or more decomposition algorithms (e.g., rectilinear polygon decomposition (RPD), etc.) and / or by using one or more machine learning modules (e.g., artificial neural networks (ANNs), random forests, decision trees, etc.) trained to decompose the superposition of two or more overlapping wireless signals into separate signal representations.

[0071] The system 200 determines and draws a bounding box around at least one of the separated signal representations. The bounding box depicts the boundary of the signal in the time and frequency domains. The bounding box marks the location of the separated signal within the spectrogram. These bounding boxes represent the separated signal representations and can be utilized along with the spectrogram to extract additional information about the separated signals to determine one or more parameters of the separated signals. This is performed by analyzing the corresponding bounding box itself, for example, by analyzing the x- and y-axis location of the bounding box, the shape of the bounding box, the color of the bounding box, the size of the bounding box, etc. Parameters that can be deduced from analyzing the bounding box include the wavelength of the corresponding separated signal, the frequency of the corresponding separated signal, the rise time of the corresponding separated signal, the fall time of the corresponding separated signal, the duration of the corresponding separated signal, the power of the corresponding separated signal, the occupied bandwidth of the corresponding separated signal, the spectral density of the corresponding separated signal, the activity profile of the corresponding separated signal, the polarization of the corresponding separated signal, etc.

[0072] In some cases, system 200 can fine-tune the generated parameters of the separated signals. System 200 achieves this by applying multiresolution time-frequency analysis, which enables system 200 to further increase the accuracy and resolution of bounding boxes around the separated signals and deduce more accurate parameter estimates. In multiresolution analysis, system 200 generates multiple additional spectrograms of the same received wireless signals, where the separated signal representations and their bounding boxes are used for the identified spectrograms. These additional spectrograms may have different resolutions. The additional spectrograms are used to deduce accurate signal parameter estimates and to fine-tune those parameters. Because limitations imposed by the Gabor uncertainty principle mean that it is not theoretically possible to accurately capture both the time and frequency characteristics of a signal using a single spectrogram, there is a trade-off between a good time-resolution representation of a signal and a good frequency-resolution representation of the same signal when analyzing a signal using a single spectrogram. System 200 generates both a good time-resolution representation of the additional spectrogram and a good frequency-resolution representation of the additional spectrogram, and uses them to fine-tune the parameters deduced from the bounding box of the separated signal. The additional spectrogram resolution depends on the signal duration and signal bandwidth and is generated to provide optimal precision alternately in the time and frequency domains. Thus, parameters related to the time domain are extracted from the high-time-resolution additional spectrogram, and parameters related to the frequency domain are extracted from the high-frequency-resolution additional spectrogram. The estimation and refinement of signal time parameters and signal frequency parameters can be performed separately by analyzing the bounding box of the separated signal in the time domain and the spectral power envelope of the signal in the frequency domain. Figure 6 shows a non-limiting example of additional spectrograms generated by system 200, where spectrogram A600-a has good time resolution.Separated signal A520-a is decomposed by system 200, and corresponding bounding boxes are drawn around them. System 200 also generates the example spectrogram shown in spectrogram B600-b, which shows an additional spectrogram with fine frequency resolution for the same received wireless signal. The same separated signal from spectrogram A600 is also shown in spectrogram B600-b as separated signal B520-b. Note that the characteristics (e.g., location, shape, etc.) of separated signal B520-b and their corresponding bounding boxes differ in the fine frequency resolution spectrogram because of the different resolution; therefore, fine-tuned frequency-related parameters can be deduced from this spectrogram. Similarly, system 200 can fine-tune time-related parameters by using the additional spectrogram with fine time resolution shown in spectrogram A600-a.

[0073] In some cases, system 200 can detect wireless signal representations in spectrograms by processing a sequence of input spectrograms continuously or periodically, frame by frame. Because multiple wireless signals are received over a given time period, their representations may appear in one or more spectrograms generated continuously or periodically over the given time period. System 200 can detect the same signal with similar or varying bounding boxes across the sequence of spectrograms. The signal bounding boxes can be processed frame by frame to deduce statistical and temporal parameters. Statistical and temporal parameters can be extracted on the detected signals. Such information can include signal activity levels, probability of occurrence and interception, power distribution, frequency stability, average duration, etc. System 200 can process signal representations continuously for each incoming spectrogram. System 200 can perform periodic processing on at least some of the spectrograms.

[0074] A non-limiting example of superposition of two or more overlapping radio signals includes the following steps, although it should be noted that at least some of the steps are optional. Downsampling the input representation. System 200 processes regions within the input spectrogram (e.g., one or more of region A 410-a, region B 410-b, or region C 410-c) as shown in FIG. 4A to generate a downsampled spectrogram, an example of which is shown in FIG. 4B. Downsampling can be achieved by using a sample-based discretization process, such as max-pooling. Downsampling can improve the signal-to-noise ratio (SNR) in the resulting downsampled spectrogram. Threshold Filtering. System 200 processes a region in the downsampled spectrogram (e.g., one or more of region A 410-a, region B 410-b, or region C 410-c) as shown in FIG. 4B to generate a threshold-filtered spectrogram, an example of which is shown in FIG. 4C. The threshold-filtered spectrogram can include only pixels having a color above a color threshold (e.g., above a color representing -40 dB). If a corresponding given pixel in the input spectrogram is below the color threshold, that pixel becomes "off" in the threshold-filtered spectrogram. System 200 can utilize pixels below the color threshold in subsequent processing steps. These pixels can be used as "off" pixels. In some cases, pixels below the color threshold can be used for image closing, for example, by filling holes in "on" pixel groups. Noise Filtering: System 200 processes a region in the threshold-filtered spectrogram (e.g., one or more of region A 410-a, region B 410-b, or region C 410-c) as shown in FIG. 4C to generate a noise-filtered spectrogram, an example of which is shown in FIG. 4D. Object Recognition. System 200 processes a region (e.g., one or more of region A 410-a, region B 410-b, or region C 410-c) in the noise-filtered spectrogram as shown in FIG. 4D to generate an object recognition spectrogram, an example of which is shown in FIG. 4E. Object recognition can be achieved, for example, by utilizing a labeling algorithm that labels connected components in the noise-filtered spectrogram to detect objects, which are contiguous groups of pixels that make up a region (e.g., one or more of region A 410-a, region B 410-b, or region C 410-c). Recognizing Overlapping Objects. System 200 processes regions (e.g., one or more of region A 410-a, region B 410-b, or region C 410-c) in the object recognition spectrogram, as shown in FIG. 4E, to generate an overlapping object recognition spectrogram, an example of which is shown in FIG. 4F. The regions are determined by identifying bounding boxes around the regions. This can be accomplished, for example, by drawing tangential polygons that contain each of the identified objects. In regions with overlapping signals (e.g., one or more of region A 410-a, region B 410-b, or region C 410-c), the tangential polygons separate the overlapping signals into separate signal representations, each represented by a corresponding bounding box around the separate signal representation, as shown, for example, in FIGS. 4G and 4H. In FIG. 4H, the separate signal representations are shown against the background of the input spectrogram representation. In the non-limiting example shown in Figures 4G and 4H, the regions, i.e., region A410-a, region B410-b, and region C410-c, are separated into separate signals, i.e., separate signal A420-a, separate signal B420-b, separate signal C420-c, and separate signal D420-d. Note that in some cases, overlapping objects have one or more intersections 430, where each intersection 430 is an area of ​​the object that is common to two or more objects. In this example, intersection 430 is common to separate signal A 420-a and separate signal B 420-b. System 200 can identify intersections 430 and associate intersections 430 with each of the objects to which intersection 430 belongs. A bounding box can be placed over all regions of the input spectrogram, as shown in FIG. 5.

[0075] The system 200 can also use the input spectrogram to optimize the signal-to-noise ratio (SNR) of the separate signal representations.

[0076] Additionally, the system 200 can employ pattern recognition techniques to determine whether the input spectrogram contains a frequency hopping signal.

[0077] System 200 may then be configured to perform one or more actions utilizing the distinct signal representations (e.g., one or more of distinct signal A 420-a, distinct signal B 420-b, distinct signal C 420-c, or distinct signal D 420-d) (block 330). Note that the actions may also be based on parameters deduced from the bounding boxes in block 320 and fine-tuned parameters determined using the multi-resolution spectrograms generated by system 200 in block 320.

[0078] The actions may include one or more of the following: Identifying the emitters that emitted the overlapping wireless signals. System 200 may have an identifier for each emitter, and each of the separated signals (e.g., one or more of distinct signal A 420-a, distinct signal B 420-b, distinct signal C 420-c, or distinct signal D 420-d) is associated with a different identified emitter. Classifying emitters that emit overlapping wireless signals. The system 200 can utilize parameters of the separated signals (e.g., one or more of distinct signal A 420-a, distinct signal B 420-b, distinct signal C 420-c, or distinct signal D 420-d) to classify emitters according to the emitted bands. For example, HF emitter, VHF emitter, UHF emitter, etc. A condensed representation of a plurality of radio signals received over time is generated. Samples of the overlapping wireless signals are extracted, thereby enabling wireless signal analysis such as locating one or more of the overlapping wireless signals, detecting one or more of the overlapping wireless signals, or accurately estimating parameters of one or more of the overlapping wireless signals. The separated signals (e.g., one or more of separate signal A 420-a, separate signal B 420-b, separate signal C 420-c, or separate signal D 420-d) are provided to one or more external systems, such as a cellular coverage planner, a radio frequency usage monitoring system, etc. The separated signals (e.g., one or more of separate signal A 420-a, separate signal B 420-b, separate signal C 420-c, or separate signal D 420-d) are provided to a user of system 200, for example, by utilizing a user interface (UI) component of system 200.

[0079] In some cases, the separated signals, their corresponding bounding boxes, and parameters may be used by different signal analysis algorithms and applications, internal or external to system 200, including spectrum monitoring, cognitive radio spectrum sensing, signal classification and clustering, communication signal demodulation and data extraction, signal direction finding and geolocation, and other algorithms and applications.

[0080] With reference to Figure 3, it should be noted that some of the blocks may be combined into an integrated block, divided into several blocks, and / or other blocks may be added. Furthermore, in some cases, the blocks may be performed in a different order than described herein. Furthermore, it should be noted that some of the blocks are optional. Also, it should be noted that while the flow charts are described in terms of the system elements that implement them, this is by no means mandatory, and the blocks may be performed by elements other than those described herein.

[0081] It is to be understood that the presently disclosed subject matter is not limited in its application to the details set forth in the description contained herein or illustrated in the drawings. The presently disclosed subject matter is capable of other embodiments and of being practiced and carried out in various ways. Accordingly, it is to be understood that the phraseology and terminology employed herein is for the purpose of description and should not be regarded as limiting. Those skilled in the art will therefore appreciate that the conception upon which the present disclosure is based may readily be utilized as a basis for the designing of other structures, methods and systems for carrying out some of the purposes of the presently disclosed subject matter.

[0082] It will also be understood that systems in accordance with the presently disclosed subject matter can be implemented, at least in part, as a suitably programmed computer. Similarly, the presently disclosed subject matter contemplates computer programs readable by a computer for performing the disclosed methods. The presently disclosed subject matter further contemplates machine-readable memory tangibly embodying a program of instructions executable by a machine for performing the disclosed methods.

Claims

1. 1. A system for time-frequency separation based on image analysis of one or more spectrograms of a plurality of radio signals received over time from one or more emitters, said system comprising: using image analysis to determine a region of the at least one spectrogram of the one or more spectrograms that includes a representation of a superposition of two or more overlapping radio signals of the radio signals, the overlapping radio signals being representations of radio signals that share a common time frequency within at least one spectrogram of the one or more spectrograms; decomposing the superposition of the two or more overlapping wireless signals within the region into separate signal representations, each given one of the separate signal representations being represented by a corresponding bounding box around the given separate signal representation; determining, for at least one distinct signal representation of the distinct signal representations, one or more parameters based on the corresponding bounding box; performing one or more actions utilizing the distinct signal representations and the parameters; a processing circuit configured to:

2. The parameters of the distinct signal representations are: (a) wavelength, (b) frequency; (c) rise time; (d) fall time; (e) duration; (f) electric power; (g) occupied bandwidth; (h) spectral density, (i) an activity profile; or (j) Polarization The system of claim 1 , comprising one or more of:

3. The processing circuitry decomposing the superposition of the two or more overlapping wireless signals to generate a plurality of additional spectrograms of the plurality of wireless signals, wherein (a) a first additional spectrogram of the additional spectrograms has a first resolution, (b) a second additional spectrogram of the additional spectrograms has a second resolution, and (c) the first resolution is different from the second resolution; For at least one given distinct signal representation of the distinct signal representations, the additional spectrograms and the corresponding parameters are utilized to determine one or more fine-tuned parameters of the given distinct signal representation, the fine-tuned parameters having a precision greater than that of the corresponding parameters. The system of claim 1 further configured to:

4. The system of claim 1 , wherein the image analysis of the one or more spectrograms of the plurality of radio signals received over time is performed continuously or periodically.

5. 5. The system of claim 4, wherein the processing circuitry, upon decomposing the superposition of the two or more overlapping radio signals, is further configured to determine one or more statistical parameters associated with at least one given distinct signal representation of the distinct signal representations based on prior occurrences of the given distinct signal representation in the one or more spectrograms.

6. The statistical parameters are (a) signal activity level; (b) probability of occurrence; (c) the probability of interception; (d) power distribution; (e) frequency stability, or (f) Average duration The system of claim 5 , comprising one or more of:

7. The act is: (a) identifying an emitter that emitted the overlapping radio signals; (b) classifying the emitters that emitted the overlapping radio signals; (c) generating a condensed representation of the plurality of wireless signals received over time; (d) extracting samples of said overlapping radio signals, thereby enabling radio signal analysis; (e) providing the overlapping radio signals to one or more external systems; or (f) providing said overlapping radio signals to users of said system. The system of claim 1 , comprising one or more of:

8. The system of claim 1 , wherein the decomposition of the superposition of the overlapping wireless signals is based on rectilinear polygon decomposition (RPD).

9. The system of claim 1 , wherein the determination of the region is also based on bands of the plurality of radio signals received over time.

10. The system of claim 1 , wherein the region is determined by identifying a bounding box around the region.

11. 1. A method for time-frequency separation based on image analysis of one or more spectrograms of a plurality of radio signals received over time from one or more emitters, said method comprising: determining, by a processing circuit, using image analysis, a region of at least one spectrogram of the one or more spectrograms that includes a superimposed representation of two or more overlapping radio signals of the radio signals, the overlapping radio signals being representations of radio signals that share a common time frequency within the at least one spectrogram of the one or more spectrograms; decomposing, by the processing circuitry, the superposition of the two or more overlapping wireless signals within the region into separate signal representations, each given one of the separate signal representations being represented by a corresponding bounding box around the given separate signal representation; determining, for at least one distinct signal representation of the distinct signal representations, one or more parameters based on the corresponding bounding box; performing, by said processing circuitry, one or more actions utilizing said distinct signal representations and said parameters; A method comprising:

12. The parameters of the distinct signal representations are: (a) wavelength, (b) frequency; (c) rise time; (d) fall time; (e) duration; (f) electric power; (g) occupied bandwidth; (h) spectral density, (i) an activity profile; or (j) Polarization The method of claim 11 , comprising one or more of:

13. generating, by the processing circuitry, a plurality of additional spectrograms of the plurality of wireless signals upon decomposing the superposition of the two or more overlapping wireless signals, wherein (a) a first additional spectrogram of the additional spectrograms has a first resolution, (b) a second additional spectrogram of the additional spectrograms has a second resolution, and (c) the first resolution is different from the second resolution; for at least one given distinct signal representation of the distinct signal representations, determining, by the processing circuitry, one or more fine-tuned parameters of the given distinct signal representation using the additional spectrogram and the corresponding parameters, the fine-tuned parameters having a precision greater than that of the corresponding parameters; The method of claim 11 further comprising:

14. The method of claim 11 , wherein the image analysis of the one or more spectrograms of the plurality of radio signals received over time is performed continuously or periodically.

15. 15. The method of claim 14, further comprising determining one or more statistical parameters associated with at least one given distinct signal representation of the distinct signal representations based on prior occurrences of the given distinct signal representation in the one or more spectrograms upon decomposing the superposition of the two or more overlapping radio signals.

16. The statistical parameters are (a) signal activity level; (b) probability of occurrence; (c) the probability of interception; (d) power distribution; (e) frequency stability, or (f) Average duration 16. The method of claim 15, comprising one or more of:

17. The act is: (a) identifying the emitter that emitted the overlapping radio signals; (b) classifying the emitters that emitted the overlapping radio signals; (c) generating a condensed representation of the plurality of wireless signals received over time; (d) extracting samples of said overlapping radio signals, thereby enabling radio signal analysis; (e) providing the overlapping radio signals to one or more external systems; or (f) providing said overlapping radio signals to a user. The method of claim 11 , comprising one or more of:

18. The method of claim 11 , wherein the determination of the region is also based on bands of the plurality of radio signals received over time.

19. The method of claim 11 , wherein the region is determined by identifying a bounding box around the region.

20. A non-transitory computer-readable storage medium having computer-readable program code embedded therein, the computer-readable program code being executable by at least one processing circuit of a computer to perform a method for time-frequency separation based on image analysis of one or more spectrograms of a plurality of radio signals received over time from one or more emitters, the method comprising: determining, by a processing circuit, using image analysis, a region of at least one spectrogram of the one or more spectrograms that includes a superimposed representation of two or more overlapping radio signals of the radio signals, the overlapping radio signals being representations of radio signals that share a common time frequency within the at least one spectrogram of the one or more spectrograms; decomposing, by the processing circuitry, the superposition of the two or more overlapping wireless signals within the region into separate signal representations, each given one of the separate signal representations being represented by a corresponding bounding box around the given separate signal representation; determining, by the processing circuitry, for at least one distinct signal representation of the distinct signal representations based on the corresponding bounding box; performing, by said processing circuitry, one or more actions utilizing said distinct signal representations and said parameters; 1. A non-transitory computer-readable storage medium comprising:

Images