Method and system for eliminating instantaneous interference based on GPU acceleration

Abstract

The present invention discloses a method and system for eliminating instantaneous interference based on GPU acceleration. The method is used to process signals received by a radio telescope, and includes the following steps: S1. Data grouping: Fourier transform the input signal, and divide the transformed frequency domain signal data into X groups by row, with each 2048 rows as a group of signal data; S2. Interference detection: Use a graphics processing unit (GPU) to detect each group of frequency domain signals to identify and mark instantaneous interference; During the detection process, set a spectral kurtosis threshold for each group of signals, and mark the channel exceeding the threshold as an instantaneous interference channel; S3. Interference elimination: Use a GPU to accelerate the instantaneous interference elimination algorithm, and eliminate the marked instantaneous interference channel; S4. Result merging: Convert the processed frequency domain signal back to the time domain through an inverse Fourier transform, and output a signal without interference. The method can improve the processing speed and retain more original data details in the process of eliminating interference.

Description

Technical Field

[0001] The present invention relates to radio physics, and in particular to a GPU-accelerated instantaneous interference elimination method and system for signal processing in radio astronomy observation. Background Art

[0002] In radio astronomy observations, radio frequency interference is a common problem that seriously affects the quality of observation data. These interferences are mixed with radio signals, making it difficult for scientists to extract effective astronomical information from them. Radio frequency interference is usually divided into two categories according to its time characteristics: long-term interference and instantaneous interference. Long-term interference refers to interference that lasts for a long time, while instantaneous interference occurs in a very short time.

[0003] Existing interference cancellation methods perform well in dealing with long-term interference, but have obvious shortcomings in dealing with instantaneous interference. When a channel experiences instantaneous interference, traditional interference cancellation methods often choose to cancel the interference of the entire channel, which may not only lead to misprocessing of non-interference data, but also may cause the loss of valuable original data. In addition, with the development of radio astronomy observation technology, the amount of data that needs to be processed has increased dramatically. When dealing with such a large data set, the processing speed of traditional central processing unit (CPU)-based interference cancellation algorithms often becomes a limiting factor. Summary of the invention

[0004] The purpose of the present invention is to provide a method and system for eliminating instantaneous interference based on GPU acceleration, which adopts an instantaneous interference elimination algorithm based on GPU acceleration and utilizes the powerful parallel computing capability of GPU to quickly and accurately identify and eliminate instantaneous interference while retaining more original data. The present invention is of great significance for improving the quality of radio astronomy observation data and accelerating data processing speed.

[0005] To achieve the above object, the present invention provides a method for eliminating instantaneous interference based on GPU acceleration, the method comprising the following steps:

[0006] S1. Data grouping: Perform Fourier transform on the input signal, and divide the transformed frequency domain signal data into X groups by row, with each 2048 rows forming a group of signal data;

[0007] S2. Interference detection: Use the graphics processing unit (GPU) to detect each group of frequency domain signals to identify and mark instantaneous interference; during the detection process, set a spectral kurtosis threshold for each group of signals, and mark the channel exceeding the threshold as an instantaneous interference channel;

[0008] S3. Interference elimination: GPU is used to accelerate the instantaneous interference elimination algorithm to eliminate the marked instantaneous interference channels;

[0009] S4. Result merging: Recombine the packet data after instantaneous interference elimination processing into a complete output signal, convert the processed frequency domain signal back to the time domain through inverse Fourier transform, and output an interference-free signal.

[0010] Furthermore, in step S1, Fourier transform is performed on the input signal to convert the signal from the time domain to the frequency domain to obtain two-dimensional frequency domain signal data.

[0011] Further, in step S2, the detection and marking of instantaneous interference are performed in the frequency domain, and are executed by using N graphics processing units (GPUs). X groups of signal data are divided into N parts, each of which includes X / N groups of signal data. The N GPUs execute simultaneously. In the processing of each group of signal data of each portion of signal data, a GPU is used to calculate the spectral kurtosis value of each column channel. When the spectral kurtosis value of a column channel exceeds a preset threshold, the column channel is marked as an instantaneous interference channel.

[0012] Furthermore, the calculation formula of the spectral kurtosis value SK is as follows:

[0013]

[0014] Among them, M is the length of the data block row, m represents a column channel, and the calculated spectral kurtosis value SK is the spectral kurtosis value of the column channel. represents the time series in the mth channel after discrete-time Fourier transform, and i represents the imaginary unit.

[0015] Further, in step S3, the interference elimination process is performed in a manner of setting the marked instantaneous interference channel to zero.

[0016] Further, in step S4, the processed signal data is merged row by row, and the merged signal is inverse Fourier transformed to convert back to the time domain and output a signal without interference.

[0017] Furthermore, the method further comprises the following steps:

[0018] S5. De-dispersion processing: De-dispersion processing is performed on the time domain data to eliminate possible dispersion effects;

[0019] S6.Data visualization: Display the processed data in the form of images.

[0020] Furthermore, the transformed frequency domain signal data is divided into 1024 groups by row.

[0021] On the other hand, the present invention provides an instantaneous interference elimination system based on GPU acceleration, the system comprising a data grouping module, an interference detection module, an interference elimination module, a result merging module, a dispersion elimination module and a data visualization module.

[0022] Beneficial effects:

[0023] Compared with the existing interference elimination algorithms, the present invention can not only effectively eliminate instantaneous interference while ensuring the processing speed, but also retain more original data details in the process of eliminating interference. This feature is particularly important for application fields that require high-precision data retention, such as radio telescope signal processing. In addition, through GPU accelerated processing, this algorithm can significantly improve the speed and efficiency of data processing, making real-time processing of large-scale data possible. This not only helps to improve the operating efficiency of astronomical observation equipment such as radio telescopes, but also provides more reliable data support for subsequent data analysis and research. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] Figure 1 Flow chart for instantaneous interference elimination;

[0025] Figure 2 It is the statistical histogram of SK value;

[0026] Figure 3 is the first original data interference graph;

[0027] Figure 4 For Figure 3 Comparison chart using existing interference elimination algorithms;

[0028] Figure 5 For Figure 3 A comparison diagram of the instantaneous interference elimination algorithm provided by the present invention;

[0029] Figure 6 is the second original data interference map;

[0030] Figure 7 For Figure 6 Comparison chart using existing interference elimination algorithms;

[0031] Figure 8 For Figure 6 Comparison diagram of the instantaneous interference elimination algorithm provided by the present invention;

[0032] Fig. 9 The timing diagram of GPU-accelerated instantaneous interference elimination operation. DETAILED DESCRIPTION

[0033] The technical solution of the present invention will be described clearly and completely below in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0034] In the description of the present invention, it should be noted that the terms "center", "upper", "lower", "left", "right", "vertical", "horizontal", "inner", "outer", etc., indicating the orientation or positional relationship, are based on the orientation or positional relationship shown in the drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as limiting the present invention. In addition, the terms "first", "second", and "third" are used for descriptive purposes only, and cannot be understood as indicating or implying relative importance.

[0035] In the description of the present invention, it should be noted that, unless otherwise clearly specified and limited, the terms "installed", "connected", and "connected" should be understood in a broad sense, for example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be a direct connection, or it can be indirectly connected through an intermediate medium, or it can be the internal communication of two components. For ordinary technicians in this field, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.

[0036] The following combination Figure 1-Figure 9 The specific embodiments of the present invention are described in detail. It should be understood that the specific embodiments described herein are only used to illustrate and explain the present invention, and are not used to limit the present invention.

[0037] The present invention provides a method for eliminating instantaneous interference based on GPU acceleration, which is used to process signals received by radio telescopes. This method provides an efficient mechanism for eliminating instantaneous interference, aiming to eliminate interference through precise processing to achieve predetermined performance indicators. On this basis, the parallel computing capability of GPU (graphics processing unit) is combined for accelerated processing, thereby achieving rapid and effective elimination of instantaneous interference in the signal. Figure 1 Specifically, the method comprises the following steps:

[0038] S1. Data grouping: Perform Fourier transform on the input signal to convert the signal from time domain to frequency domain to obtain two-dimensional frequency domain signal data. Fast Fourier Transform (Fast Fourier Transform) refers to performing FFT on the input data to convert the data from time domain to frequency domain. Data grouping processing refers to grouping the data after FFT transformation according to certain rules, and each group of data will enter the subsequent instantaneous interference elimination algorithm independently. Divide the frequency domain signal data into X groups by row, and each 2048 rows is a group of signal data. For the limit on the group length, the standard deviation of the spectral kurtosis can be used as a reference standard. For astronomical signals with Gaussian random distribution, the relationship between the standard deviation of its distribution and the length of the time series N (that is, each group in each channel contains N data points) is , the formula derivation process is as follows:

[0039] When only one channel is considered, the formula for calculating the spectral kurtosis can be rewritten into the above form without affecting the calculation results:

[0040]

[0041] Among them, m represents the mth channel, represents the time series in the mth channel after DTFT, represents the expectation of a variable, Represents the spectral kurtosis of the mth channel.

[0042] The expected value of spectral kurtosis used in the present invention is its expectation when the time series is infinitely long, and when the length of the obtained time series is limited, N, It also follows a certain distribution. It should be noted that when X satisfies the Gaussian distribution, |X|2 does not satisfy the Gaussian distribution, but here the present invention only uses the expectation and standard deviation to describe the characteristics of the distribution, while ignoring other mathematical characteristics. In the application scenario, the distribution of the spectral kurtosis near its expected value can be approximated as a Gaussian distribution. For astronomical signals with Gaussian random distribution, the following expected values ​​are calculated:

[0043]

[0044] in, represents the standard deviation of the Gaussian distribution, represents the integrand complex variable, represents the corresponding coordinates on the complex plane, satisfying , i represents the imaginary unit, represents the magnitude in the complex variable argument representation, that is, .

[0045] From this we can calculate the variance of the following two quantities:

[0046]

[0047] Among them, D represents the variance of the variable in brackets, and the meanings of other symbols are the same as above.

[0048] When the length of the obtained time series is N, the expected variance should satisfy

[0049]

[0050] Therefore, the calculated variance of the spectral kurtosis should satisfy

[0051]

[0052] The standard deviation of the distribution of spectral kurtosis is obtained as , It represents the standard deviation of the dispersion of the spectral kurtosis in the channel relative to the theoretical value and is a function of the length N of the time series.

[0053] The value of this standard deviation can be used as a characteristic deviation of the spectral kurtosis from the expected value. When considering the packet length, it should be noted that interference elimination is only effective when the spectral kurtosis dispersion of the signal is significantly smaller than the spectral kurtosis value generated by the noise. In order to effectively eliminate short-term interference, the selected time length must be long enough to reduce the dispersion. However, in order not to damage the signal in the interference-free period too much, the time length of the group needs to be as small as possible, so S(N) provides a reference indicator for the selection of time length. For example, for the self-built array in this example, the spectral kurtosis generated by the interference is basically greater than or much greater than 0.1. As long as S(N)=0.1 is guaranteed, effective separation can be achieved. Therefore, 2048 rows (S(N)= ) as a group is enough to distinguish the signal from the interference and effectively perform the interference elimination operation.

[0054] S2. Interference detection: In each set of signal data, detect and mark the channel where the instantaneous interference signal is located. This step uses multiple graphics processing units (GPUs) (2 N ) to execute, divide the X group of signal data into 2 N Each signal data includes X / 2 N Group signal data, 2 N GPUs are used to execute simultaneously to accelerate processing. If N is 3 and the number of GPUs is 8, each GPU processes 1024 / 8=128 groups of signal data. In the processing of each group of signal data of each piece of signal data, one GPU is used to calculate the spectral kurtosis value (SK value) of each column channel. The spectral kurtosis calculation formula is as follows:

[0055]

[0056] Where M is the length of the data block row, m represents a column channel, and the calculated SK value is the spectral kurtosis value of the column channel. For the above formula, a specific application process will be given, such as a (3,3) two-dimensional array [[1, 2, 3], [4, 5, 6], [7, 8, 9]]. Note that this two-dimensional data is only for the convenience of understanding the calculation process and is not real data. Calculate the SK value of the first column: =0.66 When the threshold is set to SK>0.5, the channel in this column is marked as an interference channel.

[0057] When the spectral kurtosis value of a column channel exceeds the preset threshold, the channel is marked as an instantaneous interference channel, and the interference channel is located by the spectral kurtosis method to avoid unnecessary data loss. The setting of the threshold depends on the environmental characteristics of the specific radio telescope array and can be flexibly adjusted. Taking the data collected by the array independently developed by this application as an example, the SK values ​​of non-interference data and interference data are calculated respectively, and the histogram of the SK values ​​is plotted, as shown in FIG. Figure 2 The analysis is shown in Figure 2. In the absence of interference, the SK values ​​are mainly densely distributed in a narrow range of -0.1 to 0.1; in contrast, the SK values ​​of data with interference exceed this range, including values ​​greater than 0.1 or less than -0.1. Based on the difference in SK values ​​between the two, 0.1 and -0.1 are set as preliminary thresholds to distinguish data without interference from data with interference. The selection of thresholds can also be made by using S(N) as a reference indicator. Taking the application scenario here as an example, the selected group is 2048 time points to form a time series, that is, N=2048, then the calculated standard deviation should be S(N)=0.1326, so the selected reasonable interference elimination threshold should be around 0.13. In fact, in practical applications, statistical results show that the actual spectral kurtosis dispersion is smaller than the theoretical prediction value, so it is reasonable to set the interference elimination threshold here as 0.1. When selecting other time lengths, the S(N) formula can provide a certain theoretical reference, but the actual threshold selection should be determined by the data characteristics obtained by the actual observation of the telescope according to local conditions. In order to verify the rationality of these threshold settings, the data after interference elimination is visualized. If the threshold is set too strictly (that is, too small), a large number of columns of data that are not interference may be mistakenly marked as interference channels, resulting in unnecessary elimination of non-interference data, such as Figure 4 On the contrary, if the threshold is set too loosely (i.e. too large), the channel where the real interference is located may not be effectively identified and marked, so that the interference data cannot be effectively eliminated, and the output data may be close to or even equal to the original data, such as Figure 3 As shown. Therefore, in practical applications, we need to dynamically adjust the threshold in a small range according to the data characteristics and processing effects to ensure the accuracy and effectiveness of the threshold setting. S3. Interference elimination: Interference elimination processing is performed on the marked instantaneous interference channels. Instantaneous interference elimination: Calculate the spectral kurtosis value of each column channel, mark the interference channels, and perform interference elimination processing on these interference channels. The processing method is to set the marked channels to zero.

[0058] S4. Result merging: The packet data after instantaneous interference elimination processing is recombined into a complete output signal, and the processed frequency domain signal is converted back to the time domain by inverse Fourier transform, and an interference-free signal is output. Inverse FFT transform refers to performing an inverse fast Fourier transform (IFFT) on the combined data to convert the data from the frequency domain back to the time domain. In this embodiment, the processed 1024 groups of signal data are merged by row, and the merged signal is inverse Fourier transformed to convert back to the time domain and output an interference-free signal.

[0059] In addition, the method further comprises:

[0060] Step S5. De-dispersion processing: De-dispersion processing is performed on the time domain data to eliminate possible dispersion effects.

[0061] Step S6. Data visualization: Display the processed data in the form of images for easy observation and analysis.

[0062] On the other hand, the present invention also provides a GPU-accelerated instantaneous interference elimination system, comprising:

[0063] Data grouping module: used to perform Fourier transform on the input signal, and divide the transformed frequency domain signal data into X groups by row, with each 2048 rows forming a group of signal data;

[0064] Interference detection module: used to detect each group of frequency domain signals using the graphics processing unit (GPU) to identify and mark instantaneous interference; during the detection process, a spectral kurtosis threshold is set for each group of signals, and channels exceeding the threshold are marked as instantaneous interference channels;

[0065] Interference elimination module: used to use GPU to accelerate the instantaneous interference elimination algorithm and eliminate the marked instantaneous interference channels;

[0066] Result merging module: used to recombine the packet data after instantaneous interference elimination processing into a complete output signal, convert the processed frequency domain signal back to the time domain through inverse Fourier transform, and output an interference-free signal.

[0067] The beneficial effect of the present invention is that more original data is retained while eliminating interference. By grouping the data, the scope of interference is further shortened, and then the interference channel is located for each group of data using the spectral kurtosis method. When a column in a group of data is marked as an interference channel, only the column in this group will be subjected to interference elimination processing. This process can avoid unnecessary data loss to achieve the purpose of eliminating instantaneous interference. The parallel computing power of multiple GPUs is utilized to improve the processing speed of instantaneous interference elimination. However, in the process of using GPU, the video memory must be released in time to avoid insufficient video memory causing the algorithm to be unable to perform GPU acceleration processing speed. The algorithm is implemented based on the Python language, and the program can run normally under Windows and Linux systems, with good cross-platform performance.

[0068] Figure 3-Figure 8 It intuitively demonstrates the significant improvement in performance of the instantaneous interference elimination algorithm proposed in the present invention compared with the prior art. The comparison chart consists of six sub-charts, which are described as follows:

[0069] Figure 3 and Figure 6 :Shows the interference in the first and second raw data. These interferences may be caused by a variety of external or internal factors, posing a threat to the accuracy and integrity of the data.

[0070] Figure 4 and Figure 7 :The results of applying the existing interference elimination algorithm to the first and second raw data are shown. Although the algorithm has achieved certain results in eliminating interference, it also mistakenly identifies some useful signals as interference and eliminates them, such as Figure 4 The waterfall chart in Figure 2 shows the part from line 25 to line 75. This false elimination phenomenon leads to data loss and information distortion, affecting the accuracy and reliability of subsequent data processing.

[0071] Figure 5 and Figure 8 : Demonstrates the application effect of the instantaneous interference elimination algorithm proposed in this application on the first and second original data. Figure 4 contrast, Figure 5 The waterfall diagram in Figure 1 only identifies and eliminates real interference from line 25 to line 75, while the useful signal is well preserved. This result shows that the algorithm can effectively suppress interference while maintaining the original characteristics and integrity of the data to the greatest extent, thus ensuring the accuracy and reliability of data processing.

[0072] In order to further improve the processing speed, the instantaneous interference elimination algorithm of the present invention adopts GPU acceleration technology. Fig. 9 The running time diagram of the GPU-accelerated instantaneous interference elimination algorithm is shown. Fig. 9 It can be seen that after using GPU acceleration, the processing speed is significantly improved, which is 7 times faster than the speed without acceleration. The present invention also facilitates the use of more efficient programming languages ​​for transplantation, such as C language, to give full play to its real-time processing capabilities to further improve the processing speed.

[0073] Any process or method description in the flowchart of the present invention or described in other ways herein can be understood as a module, segment or part of a code including one or more executable instructions for implementing the steps of a specific logical function or process, which can be implemented in any computer-readable medium for use by an instruction execution system, device or equipment, and the computer-readable medium can be any medium containing storage, communication, propagation or transmission programs for use by execution systems, devices or equipment, including read-only memories, magnetic disks or optical disks, etc.

[0074] In the description of this specification, the description with reference to the terms "embodiment", "example", etc. means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic representations of the above terms do not necessarily refer to the same embodiment or example. In addition, those skilled in the art can combine or combine different embodiments or examples described in this specification and the features therein without causing any contradiction.

[0075] Although the above content has shown and described the embodiments of the present invention, it can be understood that the above embodiments are exemplary and cannot be understood as limitations of the present invention. Ordinary technicians in the field can change, modify, replace, modify and other update operations on the above embodiments within the scope of the present invention.

Claims

1. The instantaneous interference elimination method based on GPU acceleration is characterized in that: The method is used to process a signal received by a radio telescope, and the method comprises the following steps: S1. Data grouping: Perform Fourier transform on the input signal, and divide the transformed frequency domain signal data into X groups by row, with each 2048 rows forming a group of signal data; S2. Interference detection: Use the graphics processing unit (GPU) to detect each group of frequency domain signals to identify and mark instantaneous interference; during the detection process, set a spectral kurtosis threshold for each group of signals, and mark the channel exceeding the threshold as an instantaneous interference channel; S3. Interference elimination: GPU is used to accelerate the instantaneous interference elimination algorithm to eliminate the marked instantaneous interference channels; S4. Result merging: recombining the packet data after instantaneous interference elimination processing into a complete output signal, converting the processed frequency domain signal back to the time domain through inverse Fourier transform, and outputting an interference-free signal; Among them, for the astronomical signal with Gaussian random distribution in S1, the relationship between the standard deviation of the distribution and the length of the time series N is calculated as follows: The formula for calculating spectral kurtosis is: Among them, m represents the mth channel, represents the time series in the mth channel after DTFT, represents the expectation of a variable, represents the spectral kurtosis of the mth channel; Calculate the expected value of: in, represents the standard deviation of the Gaussian distribution, represents the integrand complex variable, represents the corresponding coordinates on the complex plane, satisfying , i represents the imaginary unit, represents the magnitude in the complex variable argument representation, that is, ; This calculates the variance of the following two quantities: Among them, D represents the variance of the variable in brackets; When the length of the time series is N, the expected variance should satisfy: Therefore, the calculated variance of the spectral kurtosis should satisfy: The standard deviation of the distribution of spectral kurtosis is obtained as ; It represents the standard deviation of the dispersion of the spectral kurtosis in the channel relative to the theoretical value and is a function of the length N of the time series.

2. The method for eliminating instantaneous interference based on GPU acceleration according to claim 1, characterized in that: In step S1, Fourier transform is performed on the input signal to convert the signal from the time domain to the frequency domain to obtain two-dimensional frequency domain signal data.

3. The method for eliminating instantaneous interference based on GPU acceleration according to claim 1, characterized in that: In step S2, the detection and marking of instantaneous interference are performed in the frequency domain, and are executed by using N graphics processing units (GPUs). X groups of signal data are divided into N parts, each of which includes X / N groups of signal data. The N GPUs execute simultaneously. In the processing of each group of signal data of each portion of signal data, a GPU is used to calculate the spectral kurtosis value of each column channel. When the spectral kurtosis value of a column channel exceeds a preset threshold, the channel is marked as an instantaneous interference channel.

4. The method for eliminating instantaneous interference based on GPU acceleration according to claim 3, characterized in that: The calculation formula of the spectral kurtosis value SK is as follows: Where M is the length of the data block row; m represents a column channel, and the calculated spectral kurtosis value SK is the spectral kurtosis value of the column channel. represents the time series in the mth channel after discrete-time Fourier transform, and i represents the imaginary unit.

5. The method for eliminating instantaneous interference based on GPU acceleration according to claim 1, characterized in that: In step S3, the interference elimination process is performed by setting the marked instantaneous interference channel to zero.

6. The method for eliminating instantaneous interference based on GPU acceleration according to claim 1, characterized in that: In step S4, the processed signal data is merged row by row, and the merged signal is inverse Fourier transformed to convert back to the time domain and output a signal without interference.

7. The method for eliminating instantaneous interference based on GPU acceleration according to claim 1, characterized in that: The following steps are also included: S5. De-dispersion processing: De-dispersion processing is performed on the time domain data to eliminate the dispersion effect; S6.Data visualization: Display the processed data in the form of images.

8. The method for eliminating instantaneous interference based on GPU acceleration according to claim 1, characterized in that: The transformed frequency domain signal data is divided into 1024 groups by row.

9. The method for eliminating instantaneous interference based on GPU acceleration according to claim 8, characterized in that: The processed 1024 groups of signal data are merged row by row, and the merged signal is inverse Fourier transformed to convert back to the time domain and output a signal without interference.

10. GPU-accelerated instantaneous interference elimination system, characterized in that: The system is used to implement the instantaneous interference elimination method based on GPU acceleration according to any one of claims 1-9, and the system includes a data grouping module, an interference detection module, an interference elimination module, a result merging module, a dispersion elimination module and a data visualization module.

Images