CN116451030B - Baseband data pulse searching method and system based on GPU - Google Patents
Baseband data pulse searching method and system based on GPU Download PDFInfo
- Publication number
- CN116451030B CN116451030B CN202310714905.4A CN202310714905A CN116451030B CN 116451030 B CN116451030 B CN 116451030B CN 202310714905 A CN202310714905 A CN 202310714905A CN 116451030 B CN116451030 B CN 116451030B
- Authority
- CN
- China
- Prior art keywords
- pulse
- data
- result
- pulse signal
- signal
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
- 238000000034 method Methods 0.000 title claims abstract description 64
- 239000006185 dispersion Substances 0.000 claims abstract description 36
- 238000001914 filtration Methods 0.000 claims abstract description 28
- 238000012216 screening Methods 0.000 claims abstract description 7
- 238000004364 calculation method Methods 0.000 claims description 47
- 238000005070 sampling Methods 0.000 claims description 6
- 238000004088 simulation Methods 0.000 claims description 6
- 230000001131 transforming effect Effects 0.000 claims description 3
- 238000012545 processing Methods 0.000 abstract description 18
- 238000004422 calculation algorithm Methods 0.000 description 13
- 230000008569 process Effects 0.000 description 13
- 230000006870 function Effects 0.000 description 8
- 238000010586 diagram Methods 0.000 description 7
- 239000012634 fragment Substances 0.000 description 5
- 230000001427 coherent effect Effects 0.000 description 4
- 238000011161 development Methods 0.000 description 3
- 230000007246 mechanism Effects 0.000 description 3
- 230000005855 radiation Effects 0.000 description 3
- 238000005316 response function Methods 0.000 description 3
- 238000003491 array Methods 0.000 description 2
- 230000008030 elimination Effects 0.000 description 2
- 238000003379 elimination reaction Methods 0.000 description 2
- 230000006872 improvement Effects 0.000 description 2
- 239000000203 mixture Substances 0.000 description 2
- 238000010845 search algorithm Methods 0.000 description 2
- 230000003595 spectral effect Effects 0.000 description 2
- 238000012546 transfer Methods 0.000 description 2
- 230000001133 acceleration Effects 0.000 description 1
- 238000004458 analytical method Methods 0.000 description 1
- 238000004891 communication Methods 0.000 description 1
- 238000013500 data storage Methods 0.000 description 1
- 238000000354 decomposition reaction Methods 0.000 description 1
- 230000001934 delay Effects 0.000 description 1
- 238000002474 experimental method Methods 0.000 description 1
- 238000004880 explosion Methods 0.000 description 1
- 238000007781 pre-processing Methods 0.000 description 1
- 230000009467 reduction Effects 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 230000004044 response Effects 0.000 description 1
- 230000002123 temporal effect Effects 0.000 description 1
- 238000012795 verification Methods 0.000 description 1
- 230000003313 weakening effect Effects 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/10—Pre-processing; Data cleansing
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01R—MEASURING ELECTRIC VARIABLES; MEASURING MAGNETIC VARIABLES
- G01R29/00—Arrangements for measuring or indicating electric quantities not covered by groups G01R19/00 - G01R27/00
- G01R29/02—Measuring characteristics of individual pulses, e.g. deviation from pulse flatness, rise time or duration
- G01R29/027—Indicating that a pulse characteristic is either above or below a predetermined value or within or beyond a predetermined range of values
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01R—MEASURING ELECTRIC VARIABLES; MEASURING MAGNETIC VARIABLES
- G01R29/00—Arrangements for measuring or indicating electric quantities not covered by groups G01R19/00 - G01R27/00
- G01R29/02—Measuring characteristics of individual pulses, e.g. deviation from pulse flatness, rise time or duration
- G01R29/027—Indicating that a pulse characteristic is either above or below a predetermined value or within or beyond a predetermined range of values
- G01R29/0273—Indicating that a pulse characteristic is either above or below a predetermined value or within or beyond a predetermined range of values the pulse characteristic being duration, i.e. width (indicating that frequency of pulses is above or below a certain limit)
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/90—Details of database functions independent of the retrieved data types
- G06F16/903—Querying
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/22—Matching criteria, e.g. proximity measures
-
- H—ELECTRICITY
- H03—ELECTRONIC CIRCUITRY
- H03H—IMPEDANCE NETWORKS, e.g. RESONANT CIRCUITS; RESONATORS
- H03H17/00—Networks using digital techniques
- H03H17/02—Frequency selective networks
- H03H17/0248—Filters characterised by a particular frequency response or filtering method
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
- G06F2218/02—Preprocessing
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
- G06F2218/12—Classification; Matching
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02D—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN INFORMATION AND COMMUNICATION TECHNOLOGIES [ICT], I.E. INFORMATION AND COMMUNICATION TECHNOLOGIES AIMING AT THE REDUCTION OF THEIR OWN ENERGY USE
- Y02D10/00—Energy efficient computing, e.g. low power processors, power management or thermal management
Abstract
The invention discloses a baseband data pulse searching method and system based on a GPU, comprising the following steps: s1: simulating to generate a dispersion pulse signal as a matched filter; s2: performing sliding cross-correlation filtering on the matched filter and the baseband data; s3: screening effective data with pulse signal candidates through a matching result, if the matching result does not have a single pulse, indicating that the data file does not have a pulse signal, and marking the data file as an invalid data file; if the matching result has single pulse, indicating that a pulse candidate exists in the data file, and marking the data file as a valid data file; s4: analyzing the matching result of the effective data file to obtain the corresponding position of the pulse signal in the baseband data file; s5: only data segments near the pulse position are truncated as pulse candidate data. The invention realizes the real-time pulse searching and positioning with high time resolution of 1ns, reduces the data processing and storage requirements of pulse searching, and realizes the narrow pulse searching with limit of 10 ns.
Description
Technical Field
The invention relates to the astronomical field, in particular to a baseband data pulse searching method and system based on a Graphic Processing Unit (GPU).
Background
A Fast Radio Burst (FRB) is a Radio pulse signal that releases a very large amount of energy for a duration on the order of milliseconds. By 2023, 5 months, a total of 118 fast radio storms were detected. FRB can help us to better understand evolution history in universe, reveal internal mechanism of high-energy celestial phenomena such as star explosion, black hole combination, etc., and promote the development of celestial physics. For example, fast radiostorms can also be used to verify einstein's equivalence principle by comparing the time delays when photons of different energies emitted simultaneously reach an observer. However, there are still many undirected puzzles in the field of FRB research, such as the internal structure of FRB, the radiation mechanism and origin, etc.
The radio pulse signals in the universe are received by the telescope on the ground after passing through the interstellar medium, and the interstellar medium can influence the electromagnetic wave propagation group velocity of the pulse signals, so that the pulse signals are dispersive. Chromatic dispersion is embodied in the fact that the low frequency part of the pulse signal arrives at a delay from the high frequency part. The dispersion phenomenon can lead to pulse phase disorder, pulse signal profile distortion, pulse amplitude weakening and signal-to-noise ratio reduction, and is difficult to be searched and identified. It is necessary to restore the true pulse signal by the achromatic process in the course of performing the pulse search. The classical fast radio storm search flow is: firstly, preprocessing observation data, such as channel division, interference elimination and the like; secondly, carrying out achromatizing treatment on the frequency channel data to eliminate time delay caused by interstellar media; thirdly, carrying out single pulse search on the signals to obtain some candidates; and finally, screening and verifying the candidate.
There are some narrow pulses in FRB search studies that are difficult to find, and their signal-to-noise ratio is very low and therefore time resolution requirements are also higher. Because pulse width is a fundamental observation of the radiation process, and has important implications for determining burst energy, narrow pulse searching is critical for the exploration of puzzles such as fast-shot storm radiation mechanisms. Majid et al searched for a narrow pulse of 100ns width in baseband data with a time resolution of 62.5ns, while FAST, due to the data time resolution 98us constraint, searched for the narrowest pulse of FRB121102 is 0.4ms. But the time resolution of FAST baseband data is 1ns, and if high time resolution pulse searching is performed on the FAST baseband data, new narrow pulses are expected to be searched.
The high time resolution pulse search can improve the discovery efficiency and accuracy of pulse signals, but at the same time means the increase of data volume and calculation volume. The achromatizing process in the pulse search process requires a large number of floating point calculations, and the improvement of time resolution also means the improvement of the calculation time and the requirement of hardware resources. So for huge baseband data, data processing speed and memory space are one of the main obstacles to achieving high time resolution pulse search.
Disclosure of Invention
Aiming at the problems existing in the prior art, the invention aims to provide a baseband data pulse searching method and a baseband data pulse searching system based on a GPU, wherein the method can ensure 1ns time resolution and realize real-time pulse searching and positioning on huge baseband data; the method can screen out effective data fragments with pulse signals, exclude data without signals, and reduce the data volume of subsequent data processing and storage; a narrow pulse search can be achieved with a limit of 10 ns. It is another object of the present invention to provide a GPU-based baseband data pulse search system implementing the above method.
In order to achieve the above object, the present invention provides a baseband data pulse searching method based on a GPU, comprising the following steps:
step S1: simulating to generate a time domain dispersion pulse signal as a matched filter;
step S2: performing sliding cross-correlation filtering on the matched filter and the baseband data;
step S3: screening effective data with pulse signal candidates through a matching result, if the matching result does not have a single pulse, indicating that the data file does not have a pulse signal, and marking the data file as an invalid data file; if the matching result has single pulse, indicating that a pulse candidate exists in the data file, and marking the data file as a valid data file;
step S4: analyzing the matching result of the effective data file to obtain the corresponding position of the pulse signal in the baseband data file;
step S5: only data segments near the pulse position are truncated as pulse candidate data.
Further, the generation method of the matched filter comprises the following steps:
1) Simulating to generate a random pulse signal with pulse width of wt and pulse height of q;
2) Performing Fourier transform on the pulse signal, and transforming the signal from a time domain to a frequency domain;
3) Simulation generates one atWithin the frequency bandwidth, a dispersion delay chirp function with a fixed dispersion value dm, wherein vc is the center frequency and vb is the frequency bandwidth;
4) Multiplying the frequency domain pulse signal with a dispersion delay chirp function to obtain a pulse signal with superimposed dispersion delay;
5) And performing inverse Fourier transform on the frequency domain dispersion pulse signals to obtain time domain dispersion pulse signals.
Further, the pulse period T of the time domain dispersion pulse signal cannot be smaller thanMaximum of time delay within frequency bandwidth, i.e.Where D is the dispersion constant and dm is the dispersion quantity.
Further, the matched filter and the baseband data of the original signal are subjected to sliding cross-correlation calculation to realize pulse searching, the sliding step length is 1ns each time, and the sampling rates of the baseband data and the matched filter are 1ns, so that high time resolution of 1ns of pulse searching is realized.
Further, when the matched filter is overlapped with the hidden real pulse signals in the baseband data, the cross-correlation calculation result is maximum and is far greater than the cross-correlation calculation result of the noise signals in the baseband data, so that the pulse signal part is highlighted, the noise part is restrained, and the pulse searching and positioning are realized; if the obtained matched filtering result can see the prominent pulse, pulse signals exist in the baseband data; otherwise the baseband data does not have a pulse signal.
Further, the cross-correlation calculation specifically includes:
，
where f (t) and g (t) represent two signals to be cross-correlated to calculate a degree of similarity, τ represents a time offset.
Further, for the result of multiplying and adding each number of correlation kernels and original data in the cross-correlation calculation decomposition of two one-dimensional arrays, the method for splitting the correlation kernels is as follows:
length L f Correlation kernel of length L data Performing cross-correlation calculation of valid mode on the original data template of (2), and obtaining a total result with length of L result =L data -L f +1；
Splitting the related cores into n parts, wherein the length L of each part of related cores is equal to that of each other f_b Is L f N; for the relevant results of the ith part of relevant kernel, only intercept the [ i ] L of the result array f_b : i*L f_b +L result ]Partial data as valid batch results; by aligning and adding each effective batch result, a complete undivided result L is obtained result 。
Further, the method for splitting the original data template comprises the following steps: splitting the original data template, wherein the splitting is assumed to be m parts, and the length of each small part of data is L d_b The method comprises the steps of carrying out a first treatment on the surface of the Every minute data must overlap end to end (L f -1) length; and finally, splicing the related results of each piece of data to obtain a complete undivided result.
A baseband data pulse searching system based on a GPU, which is used for implementing the baseband data pulse searching method based on the GPU.
The baseband data pulse searching method and system based on the GPU can ensure the time resolution of 1ns and realize real-time pulse searching and positioning of huge baseband data; the method can screen out effective data fragments with pulse signals, exclude data without signals, and reduce the data volume of subsequent data processing and storage; a narrow pulse search can be achieved with a limit of 10 ns.
Drawings
FIG. 1 is a diagram of a matched filter frame;
FIG. 2 is a schematic diagram of a sliding correlation filter;
FIG. 3 is a flow chart of a matched filter search pulse;
FIG. 4 is a flow chart of matched filter generation;
FIG. 5 is a schematic diagram of the cross-correlation calculation principle;
FIG. 6 is a split view of a matched filter;
FIG. 7 is a split view of a data template;
FIG. 8 is a diagram of a matched filter generated by simulation;
FIG. 9 is a graph of matched filter results;
FIG. 10 is a time-frequency diagram of the result of the achromatism;
FIG. 11 is a graph of the matching result of a 3.7ms pulse width matched filter;
fig. 12 is a graph of the matching result of a 500ms pulse width matched filter.
Detailed Description
The following description of the embodiments of the present invention will be made more apparent and fully hereinafter with reference to the accompanying drawings, in which some, but not all embodiments of the invention are shown. All other embodiments, which can be made by those skilled in the art based on the embodiments of the invention without making any inventive effort, are intended to be within the scope of the invention.
In the description of the present invention, it should be noted that the directions or positional relationships indicated by the terms "center", "upper", "lower", "left", "right", "vertical", "horizontal", "inner", "outer", etc. are based on the directions or positional relationships shown in the drawings, are merely for convenience of describing the present invention and simplifying the description, and do not indicate or imply that the devices or elements referred to must have a specific orientation, be configured and operated in a specific orientation, and thus should not be construed as limiting the present invention. Furthermore, the terms "first," "second," and "third" are used for descriptive purposes only and are not to be construed as indicating or implying relative importance.
In the description of the present invention, it should be noted that, unless explicitly specified and limited otherwise, the terms "mounted," "connected," and "connected" are to be construed broadly, and may be either fixedly connected, detachably connected, or integrally connected, for example; can be mechanically or electrically connected; can be directly connected or indirectly connected through an intermediate medium, and can be communication between two elements. The specific meaning of the above terms in the present invention can be understood by those of ordinary skill in the art according to the specific circumstances.
The following describes specific embodiments of the present invention in detail with reference to the drawings. It should be understood that the detailed description and specific examples, while indicating and illustrating the invention, are not intended to limit the invention.
As shown in fig. 1 to 12, the method and the system for searching baseband data pulses based on the GPU of the present invention realize 1ns high time resolution pulse searching and positioning of huge baseband data by using simple linear calculation in the time domain. The method can realize short pulse search with the pulse width reaching the limit of 10ns for the baseband data, screen out effective data fragments with pulse signals, exclude data without signals and reduce the data volume of subsequent data processing. And the algorithm is accelerated in parallel on the graphic processor, so that the algorithm calculation is real-time, and the data storage and processing pressure of pulse search can be reduced simultaneously from two aspects of calculation speed and data quantity.
In the invention, the matched filtering is a linear filter, and the matched filtering can obtain the output with the highest signal to noise ratio. Assume that there is a certain segment of the input signal g (t) in the time domain:
g(t)=s(t)+n(t)
wherein s (t) represents the true pure signal portion; n (t) represents the noise part of the signal mixture, the power spectral density is. Designing a matched filter as shown in FIG. 1, passing the input signal through the matched filterThe wave device achieves the aim of suppressing noise and highlighting signals, so that the signal output with the highest signal-to-noise ratio is obtained.
Assuming that the impulse response function of the filter is H (t), the frequency domain transfer function is H (f), and the output signal g is obtained after the filter 0 (t) is the convolution of the input signal g (t) with the filter impulse response function h (t):
according to the convolution theorem, the convolution of a function in the time domain is equivalent to the product in the frequency domain, so that a signal part in an input signal is output as follows after passing through a filter:
the output power of the noise portion is:
at a certain sampling time t 0 The signal-to-noise ratio SNR of the corresponding output signal can be expressed as the ratio of the signal part power PS to the noise part power PN:
according to the Cauchy-Schvalz inequality, there is:
if and only ifI.e. +.>The inequality is equal to the sign of the signal to noise ratio maximum. Matched filterThe impulse response of (c) corresponds to the conjugate of the signal s (t) with a certain delay, so that the process of matching filtering the signal can be understood as the cross-correlation calculation of the signal. In the above formula: the input signal is g (t), s (t) represents the true pure signal part; n (t) represents the noise part of the signal mixture, the power spectral density isThe impulse response function of the filter is H (t), the frequency domain transfer function is H (f), and the output signal is g 0 (t), k is a constant,for a particular moment. Where the frequency domain of all parameters is denoted by capital letters and f denotes frequency.
The correlation calculation of signals is divided into autocorrelation and cross correlation, and the correlation calculation of two signals can be understood as calculating the similarity degree of two signals at a certain moment. The principle of matched filtering is applied to radio astronomical baseband data processing, so that narrow pulse signal searching is realized.
According to the characteristics of the real radio astronomical pulse signal, a pulse signal with fixed dispersion in a certain frequency range is simulated and generated as a matched filter.
The matched source slides on the original baseband data in 1ns step size to make a cross correlation calculation, and as shown in fig. 2, the correlation value of the matched filter slid to a certain position is equivalent to the overlapping area of the two signals. When the matched filter is overlapped with the hidden pulse signal part in the baseband data, the maximum value of the cross-correlation calculation result is obtained, and the cross-correlation calculation result of the noise part in the sliding correlation process is far smaller than the result of the pulse signal part, so that the mixed noise component in the baseband data is restrained, the hidden real signal part is highlighted, and the aim of searching and positioning the pulse signal is fulfilled. The specific flow of the method for searching the narrow pulse signal by matched filtering is shown in fig. 3, and the baseband data pulse searching method based on the GPU of the invention specifically comprises the following steps:
step S1: simulating to generate a dispersion pulse signal as a matched filter;
step S2: performing sliding cross-correlation filtering on the matched filter and the baseband data;
step S3: screening effective data with pulse signal candidates through the matching result, if the matching result does not have a single pulse, indicating that the data file does not have a pulse signal, and marking the data file as an invalid data file. If the matching result has single pulse, indicating that a pulse candidate exists in the data file, and marking the data file as a valid data file;
step S4: analyzing the matching result of the effective data file to obtain the corresponding position of the pulse signal in the baseband data file;
step S5: only intercepting data segments near the pulse position as pulse candidate data;
step S6: and checking and screening the pulse candidate data.
Fig. 4 shows a specific flow of matched filter generation. According to the coherent achromatism principle, a pulse signal with center frequency vc, frequency bandwidth vb, sampling rate 1ns, pulse signal period T, dispersion value dm, pulse width wt and pulse height q is simulated. Wherein the pulse period T cannot be smaller thanMaximum value of time delay within frequency bandwidth, i.e. +.>。
The generation method of the matched filter comprises the following steps:
1) Simulating to generate a random pulse signal with pulse width of wt and pulse height of q;
2) Performing Fourier transform on the pulse signal, and transforming the signal from a time domain to a frequency domain;
3) Simulation generates one atA dispersion delay chirp function having a fixed dispersion value dm within the frequency bandwidth;
4) Multiplying the frequency domain pulse signal with a dispersion delay chirp function to obtain a pulse signal with superimposed dispersion delay;
5) And performing inverse Fourier transform on the frequency domain dispersion pulse signals to obtain time domain dispersion pulse signals.
In the cross-correlation filtering process, the matched filter and the baseband data of the original signal are subjected to sliding cross-correlation calculation to realize pulse searching, and each sliding step length is 1ns.
If a pulse signal is present in the baseband data, the cross-correlation calculation result is maximized when the matched filter is overlapped with the pulse signal in the baseband data, so that a prominent pulse can be seen in the resulting matched filter result.
If no pulse signal is present in the baseband data, no prominent pulse can be seen in the resulting match result graph. And screening out effective baseband data fragments with pulse signals by judging whether the matched filtering result has pulses or not. In the process of realizing the cross-correlation calculation of the matched filter and the baseband data, the baseband data has the characteristic of large data volume, so the problem of insufficient memory can be encountered. The problem of insufficient memory is solved by splitting the matched filter and the baseband data respectively and then performing cross-correlation calculation in batches.
The cross-correlation calculation is specifically as follows:
where f (t) and g (t) represent two signals to be cross-correlated to calculate a degree of similarity, τ represents a time offset.
For example, the calculation example:。
fig. 5 is a schematic diagram of cross-correlation calculation. The nature of the cross-correlation calculation is a multiple point multiplication addition operation, which is a linear calculation. The cross-correlation calculation for two one-dimensional arrays can be decomposed into the result of multiplying each number of correlation kernels with the original data and then adding. The ith row is obtained by multiplying the ith number in the core by the data template, and the result of the ith row needs to be shifted forward by the ith number. The final correlation result of the Valid pattern is the result of adding the circled data in fig. 5 in the vertical direction. For example, the first value in the result array of FIG. 5 is。
Fig. 6 is a split view of a matched filter. When the related nucleus is split, the length is L f Correlation kernel of length L data Performing cross-correlation calculation of valid mode on the original data template of (2), and obtaining a total result with length of L result =L data -L f +1。
Splitting the related cores into n parts, wherein the length L of each part of related cores is equal to that of each other f_b Is L f And/n. For the relevant results of the ith part of relevant kernel, only intercept the [ i ] L of the result array f_b : i*L f_b +L result ]The partial data was used as an effective batch result. By aligning and adding each effective batch result, a complete undivided result L is obtained result 。
FIG. 7 is a split view of a data template. Splitting the original data template, wherein the splitting is assumed to be m parts, and the length of each small part of data is L d_b . To ensure the integrity of the result after splitting, every minute of data must overlap end to end (L f -1) length. As shown in fig. 7, the complete undivided result is finally obtained by stitching the correlation results of each small piece of data. It is noted that the data length of the original data template needs to be not smaller than the data length of the relevant core, both of which are split.
And verifying the feasibility of searching the positioning pulse:
the matched filtering method is applied to FRB121102-210 baseband data processing, and the feasibility of searching pulse signals by the matched filtering method is verified through a traditional coherent dispersion elimination algorithm. The simulation generates a pulse signal with a center frequency vc of 1325MHz, a frequency bandwidth vb of 10MHz, a sampling rate fs of 1ns, a pulse signal period T of 20.2ms, a dispersion value dm of 565.1, a pulse width of 10ns and a pulse height of 1. Fig. 8 is a matched filter generated by simulation. Fig. 9 is a graph of the matched filter results, and it can be seen that there is a significant single pulse occurrence. And positioning the position x of the pulse in the matching result, and intercepting data with the length of 0.3s near the position x of the baseband data to perform coherent achromatizing processing. Fig. 10 is a time-frequency diagram of the achromatic result, and it can be seen that there is a significant pulse signal in the frequency range of 1500-1300 MHz. The method for searching the baseband data pulse based on the GPU can accurately search out the hidden pulse signals in the FRB121102-210 baseband data and give out reasonable positioning values. The corresponding known pulse signals can be found by carrying out matched filtering on a plurality of FRB observation data, and the feasibility of the pulse signal searching method based on the matched filtering principle is verified.
Search limits for narrow pulses by the verification algorithm:
a plurality of dispersion pulse signals with different pulse widths are generated as matched filters and matched and filtered with FRB1211102-210 baseband data, and the search limit of the algorithm on the narrow pulses is verified. The actual pulse width of the known pulse baseband data 121102-210 is 3.7ms, which simulates the generation of matched filters with pulse widths of 10ns,1us,100us,3.7ms,10ms,500ms, respectively. As a result, it was found that the analog dispersion pulse signals with pulse widths of 10ns,1us,100us,3.7ms and 10ms could find a true pulse signal with pulse width of 3.7ms, whereas the matching result of 500ms had no single pulse, and could not find a true pulse signal. Fig. 11 is a graph of the matching result of the 3.7ms pwm matched filter, and fig. 12 is a graph of the matching result of the 500ms pwm matched filter. The method has the characteristic of searching for wide pulses by narrow pulses, so that narrow pulse searching with limit of 10ns can be realized by generating a matched filter with pulse width of 10 ns. The feasibility of the method for searching for narrow pulses is verified.
The data processing speed is also one of the important factors for achieving high resolution pulse search. The cross-correlation calculation in the pulse search algorithm based on the matched filtering is simple linear calculation, and the GPU can be utilized to realize the parallel acceleration of the algorithm. And realizing the matched filtering of the baseband data with different lengths and the matched filter on the GPU, and calculating the average running time of five times. From the results in table 1, it is demonstrated that the running speed of the matched filter based pulse search algorithm implemented in the GPU has enabled real-time calculations. The real-time calculation speed can effectively reduce the data processing and storage pressure in the high-resolution pulse search implementation process.
Table 1: runtime of different length data at GPU
| Data length/ms | 30 | 40 | 50 | 60 | 70 | 80 | 90 | 100 | 110 | 120 |
| Run time/ms | 22 | 27 | 28 | 32 | 37 | 36 | 42 | 46 | 77 | 48 |
| Data length/ms | 130 | 140 | 150 | 160 | 170 | 180 | 190 | 200 | 210 | 220 |
| Run time/ms | 56 | 54 | 99 | 65 | 63 | 64 | 77 | 78 | 81 | 84 |
Analysis of experimental results:
the feasibility of searching and positioning the pulse signals by the matched filtering method is verified through the experiment, and the pulse signals with the pulse width larger than the analog pulse width can be searched through setting the pulse width parameters of the matched filter, and the searching limit of the pulse signals can reach the pulse width of 10 ns. The calculation speed of the algorithm reaches the real-time level through the running time of the data with different lengths in the GPU.
According to the baseband data pulse searching method and system based on the GPU, in the process of matching filtering searching pulses, a matching source signal and baseband data which are generated in a simulating mode are all 1ns in sampling rate. The time resolution of the matched filtering method depends on the step length of sliding correlation, and the cross-correlation calculation function adopted by the algorithm sets the sliding step length to be 1ns, so that the time resolution of the matched filtering pulse signal searching method reaches 1ns, and the method can well conform to the development trend of astronomical observation of high time resolution.
Since the pulse duration of the pulse signal is short and the actual observation time is long, only a small amount of data in the large amount of observation baseband data is valid pulse signal data. The baseband data is subjected to GPU-based pulse search, effective pulse data with smaller data volume can be screened out, invalid data is discarded, the data processing and storage pressure of high-time-resolution pulse search can be reduced, and the data volume of subsequent data processing is greatly reduced. The pulse searching method based on the matched filtering is simple linear calculation, has extremely high parallelism, can accelerate the algorithm through the GPU, and realizes the real-time algorithm operation. The pulse search method based on matched filtering can reduce the data processing and storage pressure of high-resolution pulse search by both increasing the calculation speed and reducing the data volume.
Aiming at the problems of data processing and storage challenges in the process of high-time-resolution pulse search of large data volume, the invention provides a baseband data pulse search method based on GPU, which processes a large amount of baseband data by utilizing a method of matching filtering search pulses before coherent dispersion including a large amount of floating point calculation is carried out. The method realizes the searching and positioning of the pulse signals, screens out the effective data fragments with the pulse signals, deletes the data without the pulse signals, and greatly reduces the data quantity of pulse searching. And the algorithm based on the matched filtering search pulse is accelerated in parallel by using the GPU, so that calculation instantaneity is realized. The method for matching and filtering the search pulse can solve the problems of data processing and storage in the process of searching the high-time-resolution pulse from two aspects of data quantity and calculation speed, and is greatly helpful for searching and researching the high-resolution pulse. The method can reach the pulse width of 10ns for the search limit of the narrow pulse, and is helpful for searching and researching the narrow pulse signal. The time resolution of the algorithm also reaches 1ns, and the method can well conform to the development trend of high-resolution pulse search. The invention has the following advantages:
the advantages are as follows: a reduced data amount of the pulse search;
the advantages are as follows: real-time is realized, and the data processing speed is increased;
the method has the following advantages: the narrow pulse search limit reaches a pulse width of 10 ns;
the advantages are four: the temporal resolution of the algorithm reaches 1ns.
Claims (5)
1. A method for searching baseband data pulses based on a GPU, the method comprising the steps of:
step S1: simulating to generate a time domain dispersion pulse signal as a matched filter;
step S2: performing sliding cross-correlation filtering on the matched filter and the baseband data;
step S3: screening effective data with pulse signal candidates through a matching result, if the matching result does not have a single pulse, indicating that the data file does not have a pulse signal, and marking the data file as an invalid data file; if the matching result has single pulse, indicating that a pulse signal candidate exists in the data file, and marking the data file as a valid data file;
step S4: analyzing the matching result of the effective data file to obtain the corresponding position of the pulse signal in the baseband data;
step S5: only intercepting data segments near the pulse position as pulse signal candidate data;
the matched filter and the baseband data of the original signal are subjected to sliding cross-correlation calculation to realize pulse searching, the sliding step length is 1ns each time, and the sampling rates of the baseband data and the matched filter are 1ns, so that high time resolution of 1ns of pulse searching is realized;
the cross-correlation calculation is specifically as follows:
，
wherein f (t) and g (t) represent two signals to be subjected to cross-correlation calculation to compare the similarity degree, and τ represents the time offset;
the method for resolving the correlation kernel by multiplying each number of the correlation kernels by the original data and adding the multiplied result is as follows:
length L f Correlation kernel of length L data The cross-correlation calculation of valid mode is carried out on the original data template of (2), and the length of the obtained total result is L result =L data -L f +1；
Splitting the related cores into n parts, wherein the length L of each part of related cores is equal to that of each other f_b Is L f N; for the relevant results of the ith part of relevant kernel, only intercept the [ i ] L of the result array f_b : i*L f_b +L result ]Partial data as valid batch results; by aligning and adding each effective batch result, a complete undivided result L is obtained result ；
The method for splitting the original data template comprises the following steps: splitting the original data template, wherein the splitting is assumed to be m parts, and the length of each small part of data is L d_b The method comprises the steps of carrying out a first treatment on the surface of the Every small data must overlap end to end L f -1 length; and finally, splicing the related results of each piece of data to obtain a complete undivided result.
2. The GPU-based baseband data burst searching method of claim 1, wherein the matched filter generating method is:
1) Simulating to generate a random pulse signal with pulse width of wt and pulse height of q;
2) Performing Fourier transform on the pulse signal, and transforming the signal from a time domain to a frequency domain;
3) Simulation generates one atWithin the frequency bandwidth, a dispersion delay chirp function with a fixed dispersion dm is provided, wherein vc is the center frequency and vb is the frequency bandwidth;
4) Multiplying the frequency domain pulse signal with a dispersion delay chirp function to obtain a pulse signal with superimposed dispersion delay;
5) And performing inverse Fourier transform on the frequency domain dispersion pulse signals to obtain time domain dispersion pulse signals.
3. The GPU-based baseband data pulse searching method of claim 2, wherein a pulse period T of the time domain dispersive pulse signal is not less thanMaximum value of time delay within frequency bandwidth, i.e. +.>Where D is the dispersion constant and dm is the dispersion quantity.
4. The GPU-based baseband data burst searching method of claim 1, wherein when the matched filter is overlapped with a real burst signal in baseband data, a cross-correlation calculation result is maximum and is far greater than a cross-correlation calculation result of a noise signal in baseband data, so that a burst signal part is highlighted, a noise part is suppressed, and burst searching and positioning are realized; if the obtained matched filtering result can see the prominent pulse, pulse signals exist in the baseband data; otherwise the baseband data does not have a pulse signal.
5. A GPU-based baseband data burst search system for implementing a GPU-based baseband data burst search method according to any of claims 1-4.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310714905.4A CN116451030B (en) | 2023-06-16 | 2023-06-16 | Baseband data pulse searching method and system based on GPU |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310714905.4A CN116451030B (en) | 2023-06-16 | 2023-06-16 | Baseband data pulse searching method and system based on GPU |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN116451030A CN116451030A (en) | 2023-07-18 |
| CN116451030B true CN116451030B (en) | 2023-09-05 |
Family
ID=87124117
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202310714905.4A Active CN116451030B (en) | 2023-06-16 | 2023-06-16 | Baseband data pulse searching method and system based on GPU |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN116451030B (en) |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106771653A (en) * | 2016-11-25 | 2017-05-31 | 中国科学院新疆天文台 | A kind of sudden and violent real-time detection devices, systems and methods of Rapid Radio |
| CN110427878A (en) * | 2019-07-31 | 2019-11-08 | 中国科学院新疆天文台 | A kind of sudden and violent signal recognition method of Rapid Radio and system |
| CN111079608A (en) * | 2019-12-09 | 2020-04-28 | 中国科学院新疆天文台 | Quick radio storm real-time searching method |
| CN115982544A (en) * | 2023-01-09 | 2023-04-18 | 贵州师范大学 | Spark-based monopulse search method and parallelization research method thereof |
| CN116226616A (en) * | 2022-12-06 | 2023-06-06 | 之江实验室 | Scientific research data analysis platform for rapid electric storm |
-
2023
- 2023-06-16 CN CN202310714905.4A patent/CN116451030B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106771653A (en) * | 2016-11-25 | 2017-05-31 | 中国科学院新疆天文台 | A kind of sudden and violent real-time detection devices, systems and methods of Rapid Radio |
| CN110427878A (en) * | 2019-07-31 | 2019-11-08 | 中国科学院新疆天文台 | A kind of sudden and violent signal recognition method of Rapid Radio and system |
| CN111079608A (en) * | 2019-12-09 | 2020-04-28 | 中国科学院新疆天文台 | Quick radio storm real-time searching method |
| CN116226616A (en) * | 2022-12-06 | 2023-06-06 | 之江实验室 | Scientific research data analysis platform for rapid electric storm |
| CN115982544A (en) * | 2023-01-09 | 2023-04-18 | 贵州师范大学 | Spark-based monopulse search method and parallelization research method thereof |
Non-Patent Citations (1)
| Title |
|---|
| Fast Radio Burst Search: Cross Spectrum vs. Auto Spectrum Method;Lei Liu 等;arXiv;1-9 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN116451030A (en) | 2023-07-18 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Ryabov et al. | A low-noise, high-dynamic-range, digital receiver for radio astronomy applications: an efficient solution for observing radio-bursts from Jupiter, the Sun, pulsars, and other astrophysical plasmas below 30 MHz | |
| CN102608594B (en) | Method for suppressing frequency modulation stepping signal from synthesizing high resolution image grating lobe | |
| Gerbino et al. | Likelihood methods for CMB experiments | |
| CA2974134A1 (en) | Method and system for seismic data processing | |
| CN110046323B (en) | Rapid calculation method for synchronous compression transformation and reconstruction | |
| Tian et al. | A new motion parameter estimation algorithm based on SDFC-LVT | |
| CN111624556A (en) | Meteorological radar WTC (wind turbine controller) inhibition method based on morphological component analysis | |
| CN116451030B (en) | Baseband data pulse searching method and system based on GPU | |
| Murphy et al. | First results from the REAL-time Transient Acquisition backend (REALTA) at the Irish LOFAR station | |
| Bhunia et al. | Imaging-spectroscopy of a band-split type II solar radio burst with the Murchison Widefield Array | |
| Dudok de Wit et al. | First results from the SCM search‐coil magnetometer on Parker Solar Probe | |
| CN102043168B (en) | Method for carrying out simulation noise addition on digital signal | |
| Liu et al. | Speech denoising using transform domains in the presence of impulsive and gaussian noises | |
| CN106249282A (en) | A kind of frequency domain seismic channel set creation method being applicable to AVAF inverting | |
| CN115951315B (en) | Radar spoofing interference identification method and system based on improved wavelet packet energy spectrum | |
| Tian et al. | Long-time coherent integration and motion parameters estimation of radar moving target with unknown entry/departure time based on SAF-WLVT | |
| Kalmus et al. | Search method for unmodeled transient gravitational waves associated with SGR flares | |
| Wan et al. | Statistical analysis of solar radio fiber bursts and relations with flares | |
| Roberts et al. | Conditioning Jovian burst signals for passive sounding applications | |
| CN113325382A (en) | DDMA MIMO radar speed ambiguity resolution method based on global optimization phase modulation | |
| CN110988820B (en) | Design method and system of high-performance miscellaneous wave generator | |
| Novaes et al. | The BINGO project-VIII. Recovering the BAO signal in HI intensity mapping simulations | |
| CN113514798A (en) | Fourier Mellin transform-based method for estimating time difference/scale difference of broadband nonlinear frequency modulation signal | |
| Li et al. | A compressed sampling receiver based on modulated wideband converter and a parameter estimation algorithm for fractional bandlimited LFM signals | |
| Conti | Development of numerical tools for the analysis of near-field acoustic data |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |