Radio pulsar search method, system and apparatus

Abstract

The present specification provides a radio pulsar searching method, system and device, comprising: obtaining observation data of a radio telescope, and processing the observation data to generate a plurality of sets of time series data respectively corresponding to a plurality of dispersion measure channels; performing a phased periodic signal search on each set of time series data respectively, processing each set of time series data or derived data composed of each set of time series data according to the generated quasi-period to generate feature data of a plurality of candidates; inputting the feature data of the plurality of candidates into a pre-trained classification model to obtain a classification result representing the credibility of the plurality of candidates as target signals; obtaining a comprehensive score of the plurality of candidates, and sorting the plurality of candidates according to the comprehensive score to generate a pulsar candidate list.

Description

Technical Field

[0001] This specification relates to one or more embodiments in the field of astronomical signal processing technology, and in particular to a radio pulsar search method, system and device. Background Technology

[0002] Radio pulsars, as special high-density celestial objects, hold significant scientific value in fundamental physics and astrophysics. Searching for new pulsars, especially long-period pulsars, from massive amounts of radio telescope observation data is one of the core tasks in radio astronomy. Traditional pulsar search methods primarily rely on spectral analysis based on the Fast Fourier Transform (FFT). However, this method is insensitive to pulsar signals with long periods and low duty cycles, and is prone to missed detections when processing such signals, limiting its effectiveness in searching for long-period pulsars.

[0003] To address these issues, a technique called Fast Folding Algorithm (FFA) was proposed. This algorithm directly folds and superimposes signals periodically in the time domain, exhibiting a natural sensitivity to long-period signals. However, traditional FFA suffers from a fatal flaw in practical applications: to ensure search accuracy, the algorithm requires dense gridded searches within a vast parameter space (e.g., period, dispersion), leading to an exponential increase in computational cost with increasing search range and accuracy. For the massive amounts of data generated by current mainstream large-scale sky surveys, the computational resource consumption is prohibitive, rendering the algorithm impractical for large-scale applications.

[0004] On the other hand, after obtaining a large number of candidates through search algorithms, these candidates need to be screened to eliminate spurious signals caused by radio frequency interference or instrument effects. Traditionally, this process mainly relies on manual visual inspection by astronomers, which is not only time-consuming and labor-intensive, but also subject to subjective factors, becoming another bottleneck in the entire search process. In recent years, some technical solutions have proposed using artificial intelligence models such as deep learning to automatically classify the feature maps of candidate objects generated after periodic folding, replacing manual screening. However, these solutions only focus on automating the "verification" stage at the back end of the search process and do not solve the efficiency problem of the "generation" stage at the front end. That is, the problem of how to efficiently generate high-quality candidates from massive amounts of data still exists.

[0005] In summary, existing technologies are either hampered by the enormous computational demands of periodic searches or only address the automation of backend verification, lacking an end-to-end solution that deeply integrates efficient frontend search strategies with intelligent backend verification mechanisms. Therefore, providing a pulsar search method that can significantly reduce search computation costs while achieving efficient automated candidate verification is a pressing technical challenge in this field. Summary of the Invention

[0006] In view of this, the purpose of this specification is to solve the technical problems existing in the prior art, such as the huge amount of computation required for the periodic search of pulsars and the disconnect between the front-end search strategy and the back-end verification mechanism, and to provide an end-to-end solution that can deeply integrate an efficient front-end search strategy with an intelligent back-end verification mechanism.

[0007] To achieve the above objectives, one or more embodiments of this specification provide a radio pulsar search method, comprising:

[0008] Acquire observation data from a radio telescope and process the observation data to generate multiple sets of time series data corresponding to multiple dispersion channels;

[0009] For each group of time series data, a phased periodic signal search is performed. The phased periodic signal search includes: performing a first-stage search within a first period range with a first search precision to determine one or more candidate period sub-ranges; and performing a subsequent-stage search within the one or more candidate period sub-ranges with a second search precision higher than the first search precision to generate quasi-periods.

[0010] According to the quasi-period, each set of time series data or derived data composed of each set of time series data is processed to generate feature data of multiple candidates.

[0011] The feature data of the multiple candidates are input into a pre-trained classification model to obtain classification results that characterize the credibility of the multiple candidates as target signals;

[0012] A weighted calculation is performed on the multiple candidates, including one or more signal quality indicators such as signal-to-noise ratio, and the classification results to obtain a comprehensive score for the multiple candidates. The multiple candidates are then sorted according to the comprehensive score to generate a pulsar candidate list.

[0013] More preferably, the phased periodic signal search includes a first stage coarse search and a second stage fine search, wherein:

[0014] The first stage of coarse search uses a preset number of low-resolution folded bins to perform FFA search within the first period range to filter out coarse search candidates that meet the first signal-to-noise ratio threshold.

[0015] The second stage, fine search, involves performing an FFA search with a higher number of folded bins than the low resolution within a candidate period sub-range determined with the period of the coarse search candidate as the center, in order to obtain fine search candidates.

[0016] More preferably, in the second stage of fine search, the number of folded boxes with a higher resolution than the low resolution is obtained through dynamic calculation, the dynamic calculation including:

[0017] Based on the minimum period of the candidate period subrange and the sampling time of the time series data, and in conjunction with the preset physical constraints related to the sampling time of the time series data, the maximum allowable value of the folding box is determined.

[0018] Based on the numerical range of the minimum period of the candidate period subrange, a segmented strategy is used to determine the number of folded boxes for the second stage fine search within the range of the maximum allowed number of folded boxes.

[0019] More preferably, the segmentation strategy includes:

[0020] Based on the position of the minimum period of the candidate period subrange within the preset segmented period range, a preset number of folded boxes for the second stage fine search is selected.

[0021] More preferably, the step of processing each set of time series data or derived data composed of each set of time series data to generate feature data of multiple candidates includes: performing periodic folding processing on time-dispersion two-dimensional data containing multiple dispersion channels according to the quasi-period to generate phase-dispersion maps as feature data of the multiple candidates.

[0022] More preferably, the weighted calculation is based on the signal-to-noise ratio of the candidate and the classification confidence output by the deep learning model.

[0023] More preferably, the weighted calculation is a dynamic weighted calculation, which includes:

[0024] Based on the preset numerical range in which the classification confidence level falls, a corresponding weight combination is selected for weighting the signal-to-noise ratio and the classification confidence level.

[0025] More preferably, the candidate feature data is a phase-dispersion map, and the classification model is a deep learning model.

[0026] According to a second aspect of one or more embodiments of this specification, a radio pulsar search system is provided, comprising:

[0027] The data preprocessing module is used to acquire observation data from the radio telescope and process the observation data to generate multiple sets of time series data corresponding to multiple dispersion channels.

[0028] The periodic search module is used to perform a phased periodic signal search for each group of time series data. The phased periodic signal search includes: performing a first-stage search within a first period range with a first search precision to determine one or more candidate period sub-ranges, and performing a subsequent-stage search within the one or more candidate period sub-ranges with a second search precision higher than the first search precision to generate a quasi-period.

[0029] The feature generation module is used to process each set of time series data or derived data composed of each set of time series data according to the quasi-period to generate feature data of multiple candidates.

[0030] The classification verification module is used to input the feature data of the multiple candidates into a pre-trained classification model to obtain classification results that characterize the credibility of the multiple candidates as target signals;

[0031] The decision fusion module is used to perform weighted calculations based on one or more signal quality indicators, including signal-to-noise ratio, and the classification results of the multiple candidates to obtain a comprehensive score for the multiple candidates, and to sort the multiple candidates according to the comprehensive score to generate a pulsar candidate list.

[0032] According to a third aspect of one or more embodiments of this specification, an electronic device is provided, comprising: a processor; a memory for storing processor-executable instructions; wherein the processor performs the steps of the method as described in one or more of the above embodiments by executing the executable instructions.

[0033] As can be seen from the above embodiments, this specification, through one or more embodiments described above, employs a phased search strategy of "coarse search - fine search," avoiding indiscriminate rapid folding search across the entire parameter space. This allows for the rapid elimination of a large number of signal-free regions with minimal computational cost, concentrating computational resources on a few high-potential sub-ranges for fine-grained search. This significantly improves search efficiency and substantially reduces computational costs while maintaining the detection rate. Secondly, by introducing a pre-trained classification model to automatically classify and evaluate the generated candidate feature data, efficient and automated verification of candidates is achieved, replacing the time-consuming and labor-intensive manual visual inspection step in the traditional process, realizing an end-to-end automated processing flow. Finally, by fusing the candidate's signal quality indicators (such as signal-to-noise ratio) and model classification results for comprehensive evaluation, candidates can be ranked and screened more reliably, effectively suppressing misjudgments that may arise from a single indicator. While significantly improving overall search efficiency, it still maintains high sensitivity and accuracy for target signals such as long-period pulsars. Attached Figure Description

[0034] Figure 1 This is an exemplary embodiment of an architecture diagram of a radio pulsar search system.

[0035] Figure 2 This is a step of a radio pulsar search method provided in an exemplary embodiment.

[0036] Figure 3 This is a block diagram of a large model inference acceleration device provided in an exemplary embodiment.

[0037] Figure 4 This is a schematic diagram of the structure of a device provided in an exemplary embodiment. Detailed Implementation

[0038] To enable those skilled in the art to better understand the technical solutions in this specification, the technical solutions in the embodiments of this specification will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this specification, and not all embodiments. Based on the embodiments in this specification, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of this specification.

[0039] This specification provides an illustrative embodiment of an efficient radio pulsar search method and system. As an optional implementation, this scheme combines a phased periodic search strategy with an intelligent candidate verification mechanism, aiming to solve the problems of high computational cost and low efficiency of manual screening in traditional long-period pulsar searches.

[0040] See Figure 1 This diagram illustrates the architecture of a radio pulsar search system provided in this embodiment. The system can be deployed on a high-performance computing cluster or a single server and is implemented as a software system consisting of program instructions. Specifically, the system includes multiple functional modules that work together, such as: a data preprocessing module 10, a periodic search module 20, an adaptive parameter module 30, a feature generation module 40, a classification and verification module 50, and a decision fusion module 60.

[0041] The data preprocessing module 10 serves as the starting point of the data processing flow, acquiring the raw observation data from the radio telescope and converting it into a format suitable for subsequent processing. The periodic search module 20, one of the core modules of this scheme, is responsible for performing phased periodic signal searches. Internally, it can be further divided into a coarse search unit 21 for large-scale, low-precision scanning and a fine search unit 22 for small-scale, high-precision confirmation. The parameters set by the fine search unit 22, such as the number of search bins, are generated by the adaptive parameter module 30. The classification and verification module 50 intelligently identifies the candidates generated by the feature generation module 40. The decision fusion module 60 combines the signal quality information from the periodic search module 20 with the classification results from the classification and verification module 50, performing a comprehensive evaluation and ranking to output a final list of high-quality pulsar candidates.

[0042] The following will combine Figure 2 The flowchart shown in this embodiment of the application illustrates a radio pulsar search method, providing a detailed explanation of the specific workflow and implementation details of each module in the system. The method is... Figure 1 The radio pulsar search system shown is executed, and the function of each module or unit in the system is illustrated in detail in the following description.

[0043] The process begins with data preprocessing step S202, performed by data preprocessing module 10. First, the system acquires raw observation data from the radio telescope. This type of data is typically stored in a specific format, containing time-series signals recorded across multiple frequency channels. Considering the dispersion effect of pulsar signals as they traverse the interstellar medium—that is, the arrival times of signals at different frequencies differ—dispersion delay correction is necessary. In this step, data preprocessing module 10 sets a range of dispersion to be searched, for example, 0 to 1024 pc·cm. -3 Within this range, in preset step sizes (e.g., 1 pc·cm) -3 The data preprocessing module 10 iterates through the entire dispersion range. For each dispersion estimate, the module performs corresponding dispersion delay correction on the original multi-channel observation data and integrates the signals from all frequency channels to generate a one-dimensional time series. After traversing the entire dispersion range, multiple sets of time series data corresponding to different dispersion channels are generated, forming the basis for all subsequent searches and analyses. In some large-scale data processing applications, the original observation data stream can be received in real time through a message queue system such as Apache Kafka, in which case the data preprocessing module 10 can process it as a consumer program.

[0044] Furthermore, based on this set of time-series data corresponding to different dispersion channels, a time-dispersion two-dimensional dataset can be constructed. In this two-dimensional dataset, time and dispersion constitute two dimensions, and the numerical values ​​of the data points represent the signal intensity at the corresponding time and dispersion level. It should be noted that this derived data will be used in the subsequent feature generation step.

[0045] Subsequently, the process enters a phased periodic signal search stage, which is mainly executed by the periodic search module 20, and corresponds to... Figure 2 Step S204: Perform a phased periodic signal search for each group of time series data. In this embodiment, a two-stage fast folding algorithm search strategy is adopted, which includes a first-stage coarse search and a second-stage fine search. It can be understood that the core idea of ​​this strategy is to save computing resources by going "from coarse to fine". First, a low-precision coarse search is performed over a wide period range to obtain a relatively wide signal-to-noise ratio peak region P_coarse; then, a high-precision fine search is performed only within this peak region to obtain a sharp and accurately located signal-to-noise ratio peak P_fine.

[0046] Specifically, for each set of time series data generated by the data preprocessing module 10, the coarse search unit 21 performs a first-stage coarse search. The goal of this stage is to quickly scan a broad period range to discover regions where signals may exist at low cost. For example, the first period range is set to 0.01 seconds to 3 seconds. To reduce computation, a preset low-resolution number of folding bins is used, for example, set in the range of 20-25. The number of folding bins determines the resolution of the pulse profile after period folding; a lower number of bins means lower search accuracy but faster computation speed. The coarse search unit 21 uses the Fast Folding Algorithm (FFA) to process the time series and calculates the signal-to-noise ratio (SNR) for each trial period. The Fast Folding Algorithm enhances the SNR of periodic signals by repeatedly folding and superimposing the time series according to different trial periods. Then, all signals with an SNR higher than a preset first SNR threshold (e.g., SNR>3.0) are selected; these signals are called coarse search candidates. Each coarse search candidate contains a candidate period P and a corresponding dispersion value.

[0047] Next, the system determines whether coarse search candidates exist. Signals with a signal-to-noise ratio (SNR) lower than the first SNR threshold are discarded. This preserves a small number of periodic ranges with potential signal characteristics. If no coarse search candidates are found that meet the criteria, it is considered that there is no significant periodic signal in the time series data, and the process can either end early for that set of time series or proceed to the next set. If one or more coarse search candidates exist, the process proceeds to step S204, where the fine search unit 22 performs a second-stage fine search for each coarse search candidate.

[0048] Furthermore, the goal of the refined search is to perform a high-precision, small-range search around the candidate period determined by the coarse search, in order to accurately determine the period of the signal. For a coarse search candidate (assuming its period is P), it is first necessary to determine a sub-range of candidate periods for subsequent refined searches. In this embodiment, this sub-range is a small interval centered on the period P. For example, a marginal coefficient M can be set (e.g., M=0.01), then the refined search period sub-range is determined to be [P×(1-M), P×(1+M)]. For example, if the coarse search finds a candidate with a period P=1 second, then the refined search sub-range is correspondingly [0.99, 1.01] seconds.

[0049] Unlike the coarse search, which uses a fixed low resolution, the fine search uses a second search precision higher than the first search precision. In a preferred embodiment of this specification, this second search precision is achieved by using an adaptive high-resolution number of folded bins, which is dynamically calculated by the adaptive parameter module 30. This calculation process takes into account both physical constraints and the characteristics of the period itself.

[0050] First, input the minimum period_min (e.g., 0.99 seconds) of the current search period sub-interval and the sampling time tsamp (e.g., 1 millisecond) of the time series. Due to physical constraints, the pulse width cannot be less than the sampling time, therefore there is an upper limit to the number of folded bins. The adaptive parameter module 30 calculates the maximum allowed number of folded bins based on this, which can be expressed as: max_bins_allowed = periodmin / tsamp×safety_factor Here, `safety_factor` is a safety factor less than 1 (e.g., 0.9) to ensure the pulse width covers at least two sampling points, thus avoiding undersampling. In this example, `max_bins_allowed` = 0.99 / 0.001×0.9 =891.

[0051] Next, the adaptive parameter module 30 can determine the final number of bins used for fine-tuning by employing a segmented strategy based on the numerical range of period_min. For example, two period thresholds T1 (e.g., 1 second) and T2 (e.g., 20 seconds) can be set.

[0052] When period_min is less than T1, a short-period strategy is executed, which can be achieved by selecting a smaller fixed bin value, such as 200.

[0053] When period_min is between T1 and T2, a medium-period strategy is implemented, which can be a medium fixed bin value, such as 500.

[0054] When period_min is greater than T2, a long-period strategy is executed. In this case, to ensure sufficient resolution, a larger bin number, such as 1000, will be selected. It should be noted that this value should not be greater than the calculated max_bins_allowed.

[0055] In this example, since period_min is 0.99 seconds, which is less than T1, a short-period strategy is adopted, and bins=200 is selected. Subsequently, the refinement search unit 22 performs a high-precision fast folding algorithm search within the sub-range of [0.99, 1.01] seconds using 200 folding bins, and finally obtains one or more refinement search candidates with higher signal-to-noise ratios. For example, a refinement search candidate with a signal-to-noise ratio of 10.5 and a period of 1.0023 seconds is obtained.

[0056] After the search is complete, the system merges and post-processes all candidates generated by the coarse and fine searches. This processing includes: clustering multiple candidates with very similar period values ​​and retaining only the one with the highest signal-to-noise ratio (SNR) to eliminate redundancy; performing harmonic analysis to identify and remove harmonic signals whose periods are integer or fractional multiples of the main signal; and finally, filtering again based on SNR, for example, retaining all candidates with an SNR higher than 7.0. After this series of processing, the final output signal set is called a quasi-period, and these quasi-periods will serve as input for subsequent verification steps. In this example, a signal with a period of 1.0023 seconds is identified as a quasi-period.

[0057] Next, in step S206, the feature generation module 40 generates candidate object feature data for subsequent classification based on the quasi-period generated in the previous step. In this embodiment, the candidate object feature data includes a phase-dispersion map. The specific generation process is as follows: using the quasi-period (1.0023 seconds) and its corresponding dispersion (120.5 pc·cm)... -3 Centered on [115.0, 125.0] pc·cm, select a two-dimensional time-dispersion data block containing multiple dispersion channels. For example, select a dispersion range of [115.0, 125.0] pc·cm. -3 The data consists of 30 seconds of data; then, this two-dimensional data block is periodically folded using a quasi-period of 1.0023 seconds. This two-dimensional data block can be considered as derived data composed of a set of time series data. The result of the folding is a two-dimensional image, namely a phase-dispersion map. See also... Figure 3This is an example of a phase-dispersion plot. The horizontal axis of the plot represents the pulse phase (typically normalized to 0 to 1), and the vertical axis represents the dispersion. If a real pulsar signal exists, a bright signal region will appear at the corresponding real dispersion and phase in the plot. This image visually reflects the energy concentration of the signal in both phase and dispersion dimensions. For example, a phase-dispersion plot with a resolution of 64x256 pixels can be generated, where 64 corresponds to the dispersion axis and 256 corresponds to the phase axis.

[0058] Subsequently, in step S208, the classification verification module 50 analyzes the generated candidate feature data. This module embeds a pre-trained classification model. In this embodiment, the model is a deep learning model, such as a convolutional neural network (e.g., ResNet-18), specifically trained to identify pulsar signal patterns in phase-dispersion maps. The 64x256 pixel phase-dispersion map generated in the previous step is input into the model, which performs forward propagation calculations and outputs a classification result. In this embodiment, the classification result is a classification confidence score, ranging from 0 to 1, used to characterize the confidence or probability that the candidate is a real pulsar signal. Assume that after analyzing the input phase-dispersion map, the model outputs a classification confidence score of 0.92.

[0059] Finally, the process proceeds to step S210, where the decision fusion module 60 performs decision fusion and output. This step aims to combine traditional signal quality indicators with the judgment of modern artificial intelligence models to conduct a more comprehensive evaluation of the candidates. In this embodiment, the signal quality indicators include at least the signal-to-noise ratio (SNR) of the candidate, such as SNR=10.5 obtained during the refinement phase. The decision fusion module 50 first normalizes the SNR to ensure its range falls between 0 and 1 (assuming a normalized value of 0.85). Then, a weighted calculation is performed based on the candidate's SNR and the classification confidence score output by the classification model to obtain a comprehensive score. For example, the weight w of the SNR can be set. snr The weight w of the classification confidence is 0.4. conf A score of 0.6 indicates a greater emphasis on the evaluation results of the deep learning model. In this case, the comprehensive score is calculated using the formula: Score = w snr ×SNRnorm+w conf×Conf. Substituting the values, we get Score = 0.4 × 0.85 + 0.6 × 0.92 = 0.34 + 0.552 = 0.892. The system calculates such a comprehensive score for all verified candidates and sorts them in descending order based on the score, ultimately generating and outputting a list of pulsar candidates. Finally, in step S209, the system selects a portion of the top-ranked candidates (e.g., the top 100 or candidates with scores above a certain final threshold) to form the final list of pulsar candidates and outputs it. This list contains a significantly reduced number of candidates, and the proportion of true signals is effectively increased, allowing astronomers to make final confirmations or directly use it for subsequent scientific analysis.

[0060] As a preferred implementation, the weighting calculation here can be a dynamic weighting calculation. That is, the weighting coefficient w snr and w conf It is not fixed, but dynamically selected based on the preset numerical range of the classification confidence score. For example, the following rule can be set: when the classification confidence score is greater than 0.95, it indicates that the model has extremely high confidence in this candidate, and at this time, a set of weights that emphasize the model can be selected, such as w. snr =0.2, w conf =0.8; when the classification confidence is between 0.8 and 0.95, choose a balanced set of weights, such as w snr =0.5, w conf =0.5; when the classification confidence is low, more reliance can be placed on the traditional physical indicator of signal-to-noise ratio, and w can be selected. snr =0.7, w conf =0.3. This dynamic weighting strategy makes the evaluation process more flexible and intelligent.

[0061] In another illustrative embodiment provided in this specification, to further optimize search efficiency over a very large period range (e.g., from 1 second to several thousand seconds), the two-stage search is extended to a multi-stage search, such as a three-stage search. Subsequent steps such as candidate feature data generation, model validation, and decision fusion can remain consistent with the foregoing embodiments.

[0062] In a specific three-stage search implementation, the execution flow of the periodic search module 20 is as follows: The first stage involves a very coarse search. The goal of this stage is to perform a rapid scan over a very large period (e.g., 1 second to 500 seconds) using extremely low search precision. For example, a small number of folded bins, such as bins=64, can be used. The computational cost of this search is extremely low; its purpose is not precise localization, but rather to quickly eliminate most periodic regions without signal and identify several candidate periodic regions with larger widths. For example, this stage may find an initial improvement in the signal-to-noise ratio in the [28-32 seconds] and [150-160 seconds] regions.

[0063] The second phase involves a medium-precision search. For each candidate period region identified in the first phase (e.g., the [28-32 seconds] region), a medium-precision search is performed. This search precision is higher than the extremely coarse search; for example, a medium-resolution bin number, such as bins=512, is used. The goal of this search is to further improve the accuracy of localization within the already narrowed range, further reducing the candidate period range to one or more more precise sub-ranges. For example, after the medium-precision search, the candidate period might be precisely located within the narrower sub-interval of [30.1-30.3 seconds].

[0064] The third stage involves a fine-grained search. For each candidate period subrange (e.g., [30.1-30.3 seconds]) identified in the second stage, a final precise search is performed using a second search precision higher than that of the intermediate search. This stage is implemented in exactly the same way as the second-stage fine search in Embodiment 1, i.e., using a high-resolution number of folded boxes (e.g., 8192) dynamically calculated by the adaptive parameter module 30 to determine the final quasi-period.

[0065] By employing this step-by-step focusing search strategy, the system can allocate computing resources more precisely to the parameter space regions where signals are most likely to exist, thus avoiding wasting computing power in vast, ineffective regions. Compared to two-stage search, for search tasks that need to cover a very large period range, this multi-stage search strategy can significantly reduce the overall computational load and further improve efficiency.

[0066] Figure 4 This is a schematic structural diagram of a device provided in an exemplary embodiment. For example... Figure 4As shown, device 400 mainly consists of a communication interface 402, a user interface 404, a processor 406, and a data storage 408. These components are interconnected and communicate with each other via a system bus, network, or other connection mechanism 410. The communication interface 402 enables device 400 to communicate with other devices, access networks, and transmission networks via analog or digital modulation. For example, the communication interface 402 may include a chipset and antenna for wireless communication with a radio access network or access point. Furthermore, the communication interface 402 can be a wired interface such as Ethernet, Token Ring, or a USB port, or a wireless interface such as Wi-Fi, Bluetooth, Global Positioning System (GPS), or a wide-area wireless interface (e.g., WiMAX or LTE). Of course, the communication interface 402 can also support other forms of physical layer interfaces and standard or proprietary communication protocols. The communication interface 402 may also include multiple physical communication interfaces, such as Wi-Fi, Bluetooth, and wide-area wireless interfaces.

[0067] User interface 404 includes receiving user input and providing output to the user. Therefore, user interface 404 may include input components such as a keyboard, touch-sensitive or presence-sensitive panel, computer mouse, trackball, joystick, microphone, still camera, and video camera, and output components such as a display screen (which may be combined with a touch-sensitive panel), CRT, LCD, LED, display using DLP technology, printer, and other similar devices known or developed in the future. User interface 404 may also generate auditory output via speakers, speaker jacks, audio output ports, audio output devices, headphones, and other similar devices known or developed in the future. In some embodiments, user interface 404 may include software, circuitry, or other forms of logic capable of transmitting and receiving data from external user input / output devices. Additionally or alternatively, device 400 may support remote access from other devices via communication interface 402 or another physical interface (not shown). User interface 404 may be configured to receive user input, the position and movement of which may be indicated by an indicator or cursor described herein. User interface 404 may also be configured as a display device for rendering or displaying text fragments.

[0068] Processor 406 may contain one or more general-purpose processors and / or special-purpose processors.

[0069] Data storage 408 may include one or more volatile and / or non-volatile storage components and may be integrated wholly or partially with processor 406. Data storage 408 may include removable and non-removable components.

[0070] Processor 406 is capable of executing program instructions 418 (e.g., compiled or uncompiled program logic and / or machine code) stored in data storage 408 to perform the various functions described herein. Data storage 408 may comprise a non-transitory computer-readable medium on which program instructions are stored, which, when executed by device 400, enable device 400 to perform any methods, processes, or functions disclosed in this specification and / or the accompanying drawings. Processor 406 executing program instructions 418 may result in processor 406 using data 412.

[0071] For example, program instructions 418 may include an operating system 422 (e.g., an operating system kernel, device drivers, and / or other modules) installed on device 400 and one or more applications 420 (e.g., a browser, social application, or game application). Similarly, data 412 may include operating system data 416 and application data 414. Operating system data 416 is primarily accessible to the operating system 422, while application data 414 is primarily accessible to one or more applications 420. Application data 414 may reside in a file system visible or hidden from the user of device 400.

[0072] Application 420 can communicate with operating system 422 through one or more application programming interfaces (APIs). These APIs help application 420 read and / or write application data 414, transmit or receive information via communication interface 402, receive or display information on user interface 404, etc.

[0073] In some terminology, application 420 may be simply referred to as "app". Furthermore, application 420 can be downloaded to device 400 through one or more online app stores or app markets. However, applications can also be installed on device 400 in other ways, such as through a web browser or a physical interface on device 400 (e.g., a USB port).

[0074] Based on the above device structure, an exemplary embodiment of this specification also provides a radio pulsar search device, including: a processor; a memory for storing processor-executable instructions; wherein the processor executes the executable instructions to implement the steps of the method as described in any of the above embodiments.

[0075] For ease of description, the radio pulsar search system described in this specification is divided into various modules or units based on their functions. Of course, when implementing one or more of these specifications, the functions of each module or unit can be implemented in one or more software and / or hardware components, or a module that performs the same function can be implemented by combining multiple sub-modules or sub-units. For example, the division of units is merely a logical functional division; in actual implementation, there may be other division methods. For instance, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed.

[0076] What those skilled in the art will understand is:

[0077] In this specification, the terms "comprising," "including," or any other variations thereof are intended to cover a non-exclusive inclusion, such that a process, method, product, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, product, or apparatus. Without further limitation, the presence of additional identical or equivalent elements in a process, method, product, or apparatus that includes said elements is not excluded.

[0078] In this specification, “a,” “an,” and “the” do not specifically refer to the singular, but may also include the plural.

[0079] In this specification, ordinal numbers such as "first," "second," etc., do not necessarily indicate order; they are often used to distinguish between objects. For example, "first server" and "second server" usually refer to two servers. To differentiate between these two servers, they are described as "first server" and "second server." Of course, sometimes these two servers may be the same server.

[0080] In this specification, unless explicitly stated otherwise, "receiving and sending data" does not necessarily mean direct receiving and sending; it can also mean indirect receiving and sending. For example, A receiving data sent by B can be understood as A directly receiving the data sent by B, or it can be understood as A indirectly receiving the data sent by B through other entities such as C. Similarly, B sending data to A can be understood as B sending the data directly to A, or it can be understood as B indirectly sending the data to A through other entities such as C. Here, C can be one entity, or it can be two or more entities.

[0081] In this specification, unless explicitly stated otherwise, the relationships between structures can be direct or indirect. For example, when describing "A is connected to B," unless it is explicitly stated that A and B are directly connected, it should be understood that A can be directly connected to B or indirectly connected to B. Similarly, when describing "A is on top of B," unless it is explicitly stated that A is directly above B (AB is adjacent and A is above B), it should be understood that A can be directly above B or indirectly above B (AB is separated by other elements, and A is above B). And so on.

[0082] This specification uses specific terms to describe embodiments thereof. For example, "one embodiment" and / or "some embodiments" refer to a particular feature, structure, or characteristic associated with at least one embodiment of this specification. Therefore, it should be emphasized and noted that "one embodiment" or "an alternative embodiment" mentioned twice or more in different places in this specification do not necessarily refer to the same embodiment. Furthermore, those skilled in the art can combine and integrate the different embodiments or examples described in this specification, as well as the features of those different embodiments or examples, without contradiction.

[0083] Although one or more embodiments of this specification provide method steps as described in the embodiments or flowcharts, it is understood that the order of steps listed in the embodiments or flowcharts is only one of many possible execution orders and does not represent the only execution order. Therefore, when the claims involve method steps, any changes or adjustments to the order of such steps, or the parallelism between steps, are also within the scope of protection of the claims.

Claims

1. A method for searching radio pulsars, characterized in that, include: Acquire observation data from a radio telescope and process the observation data to generate multiple sets of time series data corresponding to multiple dispersion channels; A phased periodic signal search is performed on each set of time series data. The phased periodic signal search includes: a first-stage coarse search using a first search precision within a first period range to determine one or more candidate period sub-ranges; and a second-stage fine search using a second search precision higher than the first search precision within the one or more candidate period sub-ranges to generate quasi-periods. Specifically: the first-stage coarse search uses a preset low-resolution folded bin number to perform an FFA search within the first period range to filter out coarse search candidates that meet a first signal-to-noise ratio threshold; the second-stage fine search, for the coarse search candidates, uses a higher-resolution folded bin number than the low resolution within a candidate period sub-range centered on the period of the coarse search candidate to perform an FFA search to obtain fine search candidates; the coarse search candidates and the fine search candidates are then merged and filtered to obtain the quasi-periods. According to the quasi-period, each set of time series data or derived data composed of each set of time series data is processed to generate feature data of multiple candidates. The feature data of the multiple candidates are input into a pre-trained classification model to obtain classification results that characterize the credibility of the multiple candidates as target signals; A weighted calculation is performed on the multiple candidates, including one or more signal quality indicators such as signal-to-noise ratio, and the classification results to obtain a comprehensive score for the multiple candidates. The multiple candidates are then sorted according to the comprehensive score to generate a pulsar candidate list.

2. The method according to claim 1, characterized in that, In the second stage of fine search, the number of folded bins with a higher resolution than the low resolution is obtained through dynamic calculation. The dynamic calculation includes: Based on the minimum period of the candidate period subrange and the sampling time of the time series data, and in conjunction with the preset physical constraints related to the sampling time of the time series data, the maximum allowable value of the folding box is determined. Based on the numerical range of the minimum period of the candidate period subrange, a segmented strategy is used to determine the number of folded boxes for the second stage fine search within the range of the maximum allowed number of folded boxes.

3. The method according to claim 2, characterized in that, The segmentation strategy includes: Based on the position of the minimum period of the candidate period subrange within the preset segmented period range, a preset number of folded boxes for the second stage fine search is selected.

4. The method according to claim 1, characterized in that, The step of processing each set of time series data or derived data composed of each set of time series data to generate feature data of multiple candidates includes: performing periodic folding processing on time-dispersion two-dimensional data containing multiple dispersion channels according to the quasi-period to generate phase-dispersion maps as feature data of the multiple candidates.

5. The method according to claim 4, characterized in that, The weighted calculation is based on the signal-to-noise ratio of the candidate and the classification confidence score output by the classification model.

6. The method according to claim 5, characterized in that, The weighted calculation is a dynamic weighted calculation, which includes: Based on the preset numerical range in which the classification confidence level falls, a corresponding weight combination is selected for weighting the signal-to-noise ratio and the classification confidence level.

7. The method according to claim 1, characterized in that, The feature data of the multiple candidates are multiple phase-dispersion maps, the classification model is a deep learning model, and the classification result is the classification confidence output by the deep learning model.

8. A radio pulsar search system, characterized in that, include: The data preprocessing module is used to acquire observation data from the radio telescope and process the observation data to generate multiple sets of time series data corresponding to multiple dispersion channels. A periodic search module is used to perform a phased periodic signal search for each group of time series data. The phased periodic signal search includes: performing a first-stage coarse search within a first period range using a first search precision to determine one or more candidate period sub-ranges; and performing a second-stage fine search within the one or more candidate period sub-ranges using a second search precision higher than the first search precision to generate a quasi-period. Specifically: the first-stage coarse search uses a preset low-resolution folded bin number to perform an FFA search within the first period range to filter out coarse search candidates that meet a first signal-to-noise ratio threshold; the second-stage fine search, for the coarse search candidates, uses a higher-resolution folded bin number than the low resolution to perform an FFA search within a candidate period sub-range centered on the period of the coarse search candidate to obtain fine search candidates; and the coarse search candidates and the fine search candidates are merged and filtered to obtain the quasi-period. The feature generation module is used to process each set of time series data or derived data composed of each set of time series data according to the quasi-period to generate feature data of multiple candidates. The classification verification module is used to input the feature data of the multiple candidates into a pre-trained classification model to obtain classification results that characterize the credibility of the multiple candidates as target signals; The decision fusion module is used to perform weighted calculations based on one or more signal quality indicators, including signal-to-noise ratio, and the classification results of the multiple candidates to obtain a comprehensive score for the multiple candidates, and to sort the multiple candidates according to the comprehensive score to generate a pulsar candidate list.

9. A radio pulsar search device, characterized in that, include: processor; A memory for storing processor-executable instructions; wherein the processor implements the steps of the method as described in any one of claims 1-7 by executing the executable instructions.

Images