Communication signal blind identification and parameter estimation system

Abstract

The invention discloses a communication signal blind identification and parameter estimation system, which belongs to the technical field of signal analysis and comprises a receiver front end, a microwave down-conversion unit, a frequency synthesizer unit, an intermediate frequency signal processing unit, a computer processing unit and system control software. According to the method, blind identification and parameter estimation of nine communication signals in a complex electromagnetic environment can be accurately and quickly realized through design of software and hardware processing flows of the system, including AM, FM, BPSK, QPSK, OQPSK, 8PSK, FSK, MSK and QAM16, the code rate range can cover 1kHz-5MHz, the average identification accuracy can reach more than 95%, and the average estimation error of parameter estimation is less than 2%; through engineering verification, the method is easy to implement, high in speed and good in flexibility.

Description

Technical Field

[0001] The invention relates to a communication signal blind recognition and parameter estimation system, belonging to the technical field of signal analysis. Background Art

[0002] The use scenario of blind recognition of communication signals is mainly non-cooperative communication. The modulation mode and parameters of communication signals are identified and estimated without any prior information. The objects to be identified mainly come from line-fed or air-fed RF communication signals. For the recognition method, many scholars use the modulation mode recognition model method based on feature extraction. This method first extracts feature parameters from the intercepted signal, and then determines the modulation type of the signal through the trained recognition model. However, signal acquisition and preprocessing before feature extraction, as well as the selection of feature parameters, will directly affect the subsequent recognition accuracy.

[0003] Signal acquisition and preprocessing start from the analog signal input of the RF front end. If the RF input signal cannot be fully and properly processed as training data for the model, it will be difficult to train a robust model even with massive amounts of data. The amplitude, bit rate, frequency, bandwidth, sampling rate and other information contained in the signal data will directly affect the final recognition result, resulting in low accuracy and poor noise resistance in blind recognition and parameter estimation of communication signals. Summary of the invention

[0004] The purpose of the present invention is to provide a communication signal blind recognition and parameter estimation system to solve the problems of low accuracy and poor noise resistance in blind recognition and parameter estimation of communication signals in complex electromagnetic environments in the prior art.

[0005] To achieve the above objectives, the present invention is implemented by adopting the following technical solutions:

[0006] The present invention provides a communication signal blind recognition and parameter estimation system, comprising a receiver front end, a microwave down-conversion unit, a frequency synthesis unit, an intermediate frequency signal processing unit, a computer processing unit and system control software;

[0007] The receiver front end is used to: receive the RF analog signal and input it into the microwave down-conversion unit;

[0008] The frequency synthesizer unit is used to: generate a local oscillator signal and input it into the microwave down-conversion unit;

[0009] The microwave down-conversion unit is used to: pre-process the RF analog signal, mix it with the local oscillator signal, obtain the intermediate frequency analog signal after low-pass filtering, and input it into the intermediate frequency signal processing unit;

[0010] The intermediate frequency signal processing unit is used to: convert the intermediate frequency analog signal into an intermediate frequency digital signal, perform frequency and rate processing on the intermediate frequency digital signal according to the control parameters sent by the computer processing unit, obtain a low frequency digital signal and send it to the computer processing unit;

[0011] The computer processing unit is used to: calculate the control parameters according to the control instructions issued by the system control software and send them to the intermediate frequency signal processing unit, and send the low frequency digital signal to the system control software;

[0012] The system control software is used to generate and issue control instructions, perform blind identification and parameter estimation on low-frequency digital signals, and obtain modulation type and parameter estimation results.

[0013] Furthermore, the preprocessing performed by the microwave down-conversion unit on the radio frequency analog signal includes limiting amplification and automatic gain control.

[0014] Furthermore, the intermediate frequency signal processing unit includes an A / D sampling module and an FPGA digital signal processing module;

[0015] The A / D sampling module is used to: sample the intermediate frequency analog signal to obtain an intermediate frequency digital signal, and input the intermediate frequency digital signal into the FPGA digital signal processing module;

[0016] The FPGA digital signal processing module is used to: receive the control parameters, including the frequency word and the decimation multiple, down-convert and down-sample the intermediate frequency digital signal according to the control parameters to obtain a low-frequency digital signal, and send the low-frequency digital signal to the computer processing unit.

[0017] Furthermore, the intermediate frequency digital signal is high-rate multi-channel data, and the low-frequency digital signal is low-rate single-channel data.

[0018] Furthermore, the FPGA digital signal processing module and the computer processing unit communicate with each other using the SRIO protocol, and the computer processing unit and the system control software communicate with each other using Ethernet.

[0019] Furthermore, the system control software includes display control software and an identification module;

[0020] The display and control software is used to: issue control instructions according to the frequency band and bandwidth of the radio frequency analog signal, the control instructions include frequency and sampling rate, receive the modulation type and parameter estimation results output by the recognition module and display them on the interface;

[0021] The identification module is used to: obtain a signal to be identified by preprocessing the low-frequency digital signal in the frequency domain, extract multidimensional statistical features of the signal to be identified, input the multidimensional statistical features into a pre-trained modulation type identification model for prediction to obtain the modulation type, and estimate the modulation parameters of the signal to be identified according to the predicted modulation type to obtain a parameter estimation result.

[0022] Furthermore, the frequency domain preprocessing includes removing the DC component and accurately estimating and down-converting the carrier frequency.

[0023] Furthermore, the multidimensional statistical features include time domain features, amplitude features, phase features and frequency domain features;

[0024] The time domain characteristics include fourth-order cumulative quantities and sixth-order cumulative quantities;

[0025] The amplitude characteristics include the compactness of the zero-centered normalized instantaneous amplitude, the standard deviation of the absolute value of the zero-centered normalized instantaneous amplitude, and the standard deviation of the zero-centered normalized non-weak signal segment instantaneous amplitude;

[0026] The phase feature includes the standard deviation of the zero-center normalized non-weak signal segment instantaneous phase nonlinear component and the standard deviation of the absolute value of the zero-center normalized non-weak signal segment instantaneous phase nonlinear component;

[0027] The frequency domain features include the zero-centered normalized instantaneous frequency compactness, the standard deviation of the absolute value of the zero-centered normalized non-weak signal instantaneous frequency, the proportion of the top points of the power spectrum, and the standard deviation of the top point set of the power spectrum.

[0028] Furthermore, the modulation type recognition model is a random forest model;

[0029] The training of the modulation type recognition model includes:

[0030] Constructing a training set based on the multidimensional statistical features extracted from the historically acquired low-frequency digital signals after the frequency domain preprocessing;

[0031] The training set is input into the modulation type recognition model, and the trained modulation type recognition model is obtained through iterative training.

[0032] Further, estimating the modulation parameters of the signal to be identified according to the predicted modulation type to obtain a parameter estimation result includes:

[0033] If the predicted modulation type is AM, the signal to be identified is an AM signal, and the AM carrier frequency is calculated by the following method: the spectrum of the AM signal is calculated, the threshold of the AM carrier frequency is set according to the amplitude of the spectrum, the carrier frequencies at both ends of the AM signal spectrum are obtained, and the symmetry of the AM carrier frequency is used to calculate the average distance from the carrier frequencies at both ends of the AM signal spectrum to the center frequency point, and the AM carrier frequency is calculated by formula (1):

[0034] F AM =D* f s / n(1);

[0035] Among them, f AM is the AM carrier frequency, D is the average distance from the carrier frequency at both ends of the AM signal spectrum to the center frequency, and f s is the sampling rate, n is the number of signal points;

[0036] The AM depth is calculated by the following method: a middle section of the AM signal is selected to calculate the modulus of the complex signal, and a ratio of the difference between the maximum values ​​of the modulus and the sum of the maximum values ​​is calculated, where the ratio is the AM depth;

[0037] If the predicted modulation type is FM, the signal to be identified is an FM signal, and the FM frequency deviation is calculated by the following method: the spectrum of the FM signal is calculated, and the threshold of the FM frequency deviation is set according to the amplitude of the spectrum, and the frequencies at the two ends of the FM spectrum above the threshold are obtained. The symmetry of the FM spectrum is used to calculate the average distance from the two ends of the FM signal spectrum to the center frequency point, and the FM frequency deviation is calculated by formula (2):

[0038] F FM =D1* f s / n(2);

[0039] Among them, F FM is the FM frequency deviation, D1 is the average distance from the frequencies at the two ends of the FM signal spectrum to the center frequency;

[0040] The FM carrier frequency is calculated by the following method: the center frequency of the FM signal is accurately estimated to obtain the accurately estimated carrier frequency, the FM signal is down-converted using the accurately estimated carrier frequency, and its instantaneous phase sequence is calculated modulo 2π, the value range of the instantaneous phase sequence is (-π, π), and the instantaneous phase sequence is a folded phase sequence. The non-folded phase sequence is recovered from the instantaneous phase sequence according to the corrected phase sequence, and then the instantaneous frequency sequence is estimated according to the non-folded phase sequence. Finally, the spectrum of the instantaneous frequency sequence is calculated, and the frequency point corresponding to the maximum value of the spectrum amplitude of the instantaneous frequency sequence is the FM carrier frequency;

[0041] The modified phase sequence is calculated by formula (3):

[0042] (3);

[0043] Where C(i) is the corrected phase sequence corresponding to the i-th signal point (i=1,…,n), and θ(i) is the instantaneous phase sequence corresponding to the i-th signal point;

[0044] The method of recovering the non-folding phase sequence from the instantaneous phase sequence according to the modified phase sequence is performed by formula (4):

[0045] Ф(i)=θ(i)+ C(i-1)(4)

[0046] Where Ф(i) is the unfolded phase sequence corresponding to the i-th signal point;

[0047] The instantaneous frequency sequence of the unfolded phase sequence is calculated by formula (5):

[0048] (5);

[0049] Among them, f(i) is the instantaneous frequency sequence corresponding to the i-th signal point, T s is the sampling period, T s =1 / f s ;

[0050] If the predicted modulation type is FSK, the signal to be identified is an FSK signal. The FSK frequency deviation is calculated by the following method: the spectrum of the FSK signal is calculated, and the threshold of the FSK frequency deviation is set according to the amplitude of the spectrum. The frequencies at the two ends of the FSK signal spectrum above the threshold are obtained. The symmetry of the FSK spectrum is used to calculate the average distance from the impact at both ends of the FSK signal spectrum to the center frequency point. The FSK frequency deviation is calculated by formula (6):

[0051] F FSK =D2* f s / n(6);

[0052] Among them, f FSK is the FSK frequency deviation, and D2 is the average distance from the impulses at both ends of the FSK signal spectrum to the center frequency point;

[0053] The symbol rate of FSK / BPSK / QPSK / OQPSK / 8PSK / MSK / QAM16 is calculated by the following method: Calculate the bandwidth of the signal to be identified, according to the formula T=1 / B W Calculate the symbol period, B W is the bandwidth. The signal to be identified is delayed by half a symbol period and then multiplied with the original signal to be identified to obtain a delayed multiplication sequence. The delayed multiplication sequence is subjected to periodic spectrum line detection, and the average period of the delayed multiplication sequence is calculated based on the result of periodic spectrum line detection. Based on the average period of the delayed multiplication sequence, the symbol rate is calculated using formula (7):

[0054] R b =T'*f s / n(7);

[0055] Among them, R b is the symbol rate, and T' is the average period of the delay sequence.

[0056] Compared with the prior art, the beneficial effects achieved by the present invention are:

[0057] The invention provides a communication signal blind recognition and parameter estimation system, which provides a complete signal processing flow from the radio frequency front end to the zero-frequency back end, and is easy to implement, fast and flexible. By designing the system software and hardware processing flow, the signal has significant characteristics after being processed layer by layer, and the training data required for the modulation type recognition model is less. The trained model has high recognition accuracy, good noise resistance and strong generalization ability. This signal processing flow and accurate recognition estimation also make the parameter estimation result more accurate.

[0058] This design scheme of the present application sends the sampling rate as a control parameter to the FPGA digital signal processing module. There are two advantages of completing downsampling in the FPGA digital signal processing module: First, when the number of sampling points is constant, the sampling rate is adapted to the bit rate, which is convenient for subsequent extraction of signal features; Second, the powerful parallel processing capability of the FPGA digital signal processing module is used to efficiently solve the conversion of signal frequency and rate, thereby improving the overall speed of the system. In the recognition module, signal processing is performed based on C++, and a simple and practical algorithm is designed in combination with the characteristics of noise, power changes, and signal modulation types in actual scenarios, further improving the system's computing speed. BRIEF DESCRIPTION OF THE DRAWINGS

[0059] Figure 1 is a schematic diagram of a communication signal blind recognition and parameter estimation system provided by an embodiment of the present invention;

[0060] Figure 2 It is a processing flow chart of the FPGA digital signal processing module provided by an embodiment of the present invention;

[0061] Figure 3 is a processing flow chart of the identification module provided by an embodiment of the present invention;

[0062] Figure 4 is a flowchart of symbol rate estimation provided by an embodiment of the present invention;

[0063] Figure 5 This is a code element period spectrum line detection diagram provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0064] The present invention will be further described below in conjunction with the accompanying drawings. The following embodiments are only used to more clearly illustrate the technical solutions of the present invention and are not intended to limit the protection scope of the present invention.

[0065] Example 1.

[0066] The present invention provides a communication signal blind recognition and parameter estimation system, comprising a receiver front end, a microwave down-conversion unit, a frequency synthesis unit, an intermediate frequency signal processing unit, a computer processing unit and system control software;

[0067] The receiver front end is used to: receive the RF analog signal and input it into the microwave down-conversion unit;

[0068] The frequency synthesizer unit is used to: generate a local oscillator signal and input it into the microwave down-conversion unit;

[0069] The microwave down-conversion unit is used to: pre-process the RF analog signal, mix it with the local oscillator signal, obtain the intermediate frequency analog signal after low-pass filtering, and input it into the intermediate frequency signal processing unit;

[0070] The intermediate frequency signal processing unit is used to: convert the intermediate frequency analog signal into an intermediate frequency digital signal, perform frequency and rate processing on the intermediate frequency digital signal according to the control parameters sent by the computer processing unit, obtain a low frequency digital signal and send it to the computer processing unit;

[0071] The computer processing unit is used to: calculate the control parameters according to the control instructions issued by the system control software and send them to the intermediate frequency signal processing unit, and send the low frequency digital signal to the system control software;

[0072] The system control software is used to generate control instructions and send them to perform blind identification and parameter estimation on low-frequency digital signals to obtain modulation type and parameter estimation results.

[0073] The present invention designs control instructions and software and hardware processing flows so that the modulation type recognition model requires less training data, the trained model has high accuracy and good robustness, and thus improves the accuracy of communication signal analysis.

[0074] Example 2.

[0075] like Figure 1 As shown, the present invention provides a communication signal blind recognition and parameter estimation system, including a receiver front end, a microwave down-conversion unit, a frequency synthesis unit, an intermediate frequency signal processing unit, a computer processing unit and system control software.

[0076] like Figure 1 As shown in the data flow, the RF analog signal S from the empty feed or the first feed RF The frequency synthesizer unit is used to generate the local oscillator signal required for frequency conversion, and the microwave down-conversion unit converts the input RF analog signal SRF The limited amplification and automatic gain control are performed, and then the local oscillator signal generated by the frequency synthesizer unit is mixed and down-converted to obtain the intermediate frequency analog signal S MF .

[0077] like Figure 1 As shown in the control flow and data flow in the figure, the intermediate frequency signal processing unit includes an A / D sampling module and an FPGA digital signal processing module. The A / D sampling module can process the intermediate frequency analog signal S MF Sampling is performed to obtain an intermediate frequency digital signal S0. The sampled intermediate frequency digital signal S0 is high-speed multi-channel data and is input into the FPGA digital signal processing module. The FPGA digital signal processing module down-converts the input signal S0 to near zero frequency according to the control parameters sent by the computer unit, and obtains a low-frequency digital signal S1 after down-sampling.

[0078] FPGA, the full English name is Field Programmable Gate Array, and its Chinese name is field programmable logic gate array.

[0079] like Figure 1 As shown in the control flow and data flow in the figure, the computer processing unit can perform real-time calculations on the system, receive control instructions (frequency and sampling rate) issued by the system control software, calculate the control parameters (frequency word and decimation multiple) of the FPGA digital signal processing module according to the control instructions, and then send them to the FPGA digital signal processing module through the SRIO (Serial Rapid I / O, high-speed interconnection technology) protocol; receive data (low-frequency digital signal) fed back by the FPGA digital signal processing module, and then send it to the system control software.

[0080] like Figure 2 As shown in the figure, the specific processing flow of the FPGA digital signal processing module is as follows: after the high sampling rate multi-channel intermediate frequency digital signal S0 output by the A / D acquisition module is input into the FPGA digital signal processing module, the DDS module is used to generate corresponding data according to the received frequency word, and the signal S0 is multiplied to obtain a signal S down-converted to near zero frequency. 01 Then, the multi-channel signal is transformed into one channel through the extraction and filtering module to obtain a single-channel signal S 02 , and then use the CIC filter module to downsample according to the received decimation multiple to obtain a low-frequency single-channel low-sampling signal S1 (low-frequency digital signal), and finally output the signal S1 data to the computer processing unit using the SRIO protocol.

[0081] like Figure 1As shown in the control flow and data flow in , the system control software includes display control software and identification module. The display control software can detect the signal spectrum in the frequency band in real time. When issuing control instructions, the frequency is set according to the frequency band of the detected signal to be identified, and the sampling rate is set according to the bandwidth of the signal to ensure that the signal can be collected at an appropriate sampling rate. The identification module can blindly identify and estimate the parameters of the received signal, and feed back the modulation type and related modulation parameters to the display control software.

[0082] like Figure 3 As shown, the specific processing flow of the recognition module is: after receiving the data fed back by the computer processing unit, the data is parsed to obtain the low-frequency digital signal S1 output by the FPGA digital signal processing module, the DC component of the signal S1 is removed, and the carrier frequency is accurately estimated to obtain the signal to be identified S2, and then the multi-dimensional statistical features of the signal to be identified S2 are extracted, and the feature parameters are input into the pre-trained random forest recognition model. The recognition model selects the modulation type with the highest probability from the candidate modulation type set as the prediction result, and estimates the relevant parameters of the signal to be identified S2 according to the predicted modulation type, and finally sends the modulation type recognition result and the parameter estimation result to the display and control software.

[0083] Figure 3 The multi-dimensional statistical features in the identification module processing flow shown include time domain features, amplitude features, phase features and frequency domain features;

[0084] The time domain characteristics include fourth-order cumulants and sixth-order cumulants;

[0085] The amplitude characteristics include the compactness of the zero-centered normalized instantaneous amplitude, the standard deviation of the absolute value of the zero-centered normalized instantaneous amplitude, and the standard deviation of the zero-centered normalized instantaneous amplitude of the non-weak signal segment;

[0086] The phase characteristics include the standard deviation of the instantaneous phase nonlinear component of the zero-center normalized non-weak signal segment and the standard deviation of the absolute value of the instantaneous phase nonlinear component of the zero-center normalized non-weak signal segment;

[0087] The frequency domain features include the compactness of the zero-centered normalized instantaneous frequency, the standard deviation of the absolute value of the zero-centered normalized instantaneous frequency of non-weak signals, the proportion of the top points of the power spectrum, and the standard deviation of the set of top points of the power spectrum.

[0088] Figure 3 The modulation type recognition model in the identification module processing flow shown is a random forest model.

[0089] The training of the modulation type recognition model includes:

[0090] Constructing a training set based on multi-dimensional statistical features extracted from historically acquired low-frequency digital signals;

[0091] The training set is input into the modulation type recognition model, and a trained modulation type recognition model is obtained through iterative training.

[0092] Figure 3 The estimation process of the modulation parameters in the identification module processing flow shown includes:

[0093] If the predicted modulation type is AM, the signal to be identified is an AM signal. The AM carrier frequency is calculated by the following method: the spectrum of the AM signal is calculated, the threshold of the AM carrier frequency is set according to the amplitude of the spectrum, and the carrier frequencies at both ends of the AM signal spectrum are obtained. The symmetry of the AM carrier frequency is used to calculate the average distance from the carrier frequencies at both ends of the AM signal spectrum to the center frequency point. The AM carrier frequency is calculated by formula (1):

[0094] F AM =D* f s / n(1);

[0095] Among them, f AM is the AM carrier frequency, D is the average distance from the carrier frequency at both ends of the AM signal spectrum to the center frequency, and f s is the sampling rate, and n is the number of signal points.

[0096] The AM depth is calculated by the following method: select the middle section of the AM signal to calculate the modulus of its complex signal, and calculate the ratio of the difference between the maximum values ​​of the modulus and the sum of the maximum values. The ratio is the AM depth.

[0097] If the predicted modulation type is FM, the signal to be identified is an FM signal. The FM frequency deviation is calculated by the following method: the spectrum of the FM signal is calculated, and the threshold of the FM frequency deviation is set according to the amplitude of the spectrum. The frequencies at the two ends of the FM spectrum above the threshold are obtained. The symmetry of the FM spectrum is used to calculate the average distance from the two ends of the FM signal spectrum to the center frequency point. The FM frequency deviation is calculated by formula (2):

[0098] F FM =D1* f s / n(2);

[0099] Among them, F FM is the FM frequency deviation, and D1 is the average distance from the frequencies at the two ends of the FM signal spectrum to the center frequency.

[0100] The FM carrier frequency is calculated by the following method: the center frequency of the FM signal is accurately estimated to obtain the accurately estimated carrier frequency, the FM signal is down-converted using the accurately estimated carrier frequency, and its instantaneous phase sequence is calculated modulo 2π, the value range of the instantaneous phase sequence is (-π, π), and the instantaneous phase sequence is a folded phase sequence. The non-folded phase sequence is recovered from the instantaneous phase sequence according to the corrected phase sequence, and then the instantaneous frequency sequence is estimated according to the non-folded phase sequence. Finally, the spectrum of the instantaneous frequency sequence is calculated, and the frequency point corresponding to the maximum value of the spectrum amplitude of the instantaneous frequency sequence is the FM carrier frequency;

[0101] The modified phase sequence is calculated by formula (3):

[0102] (3);

[0103] Where C(i) is the corrected phase sequence corresponding to the i-th signal point (i=1,…,n), and θ(i) is the instantaneous phase sequence corresponding to the i-th signal point;

[0104] The method of recovering the non-folding phase sequence from the instantaneous phase sequence according to the modified phase sequence is performed by formula (4):

[0105] Ф(i)=θ(i)+ C(i-1)(4)

[0106] Where Ф(i) is the unfolded phase sequence corresponding to the i-th signal point;

[0107] The instantaneous frequency sequence of the unfolded phase sequence is calculated by formula (5):

[0108] (5);

[0109] Among them, f(i) is the instantaneous frequency sequence corresponding to the i-th signal point, T s is the sampling period, T s =1 / f s ;

[0110] If the predicted modulation type is FSK, the signal to be identified is an FSK signal. The FSK frequency deviation is calculated by the following method: the spectrum of the FSK signal is calculated, and the threshold of the FSK frequency deviation is set according to the amplitude of the spectrum. The frequencies at the two ends of the FSK signal spectrum above the threshold are obtained. The symmetry of the FSK spectrum is used to calculate the average distance from the impact at both ends of the FSK signal spectrum to the center frequency point. The FSK frequency deviation is calculated by formula (6):

[0111] F FSK =D2* f s / n(6);

[0112] Among them, f FSKis the FSK frequency deviation, and D2 is the average distance from the impact at both ends of the FSK signal spectrum to the center frequency point.

[0113] like Figure 4 As shown, the symbol rate of FSK / BPSK / QPSK / OQPSK / 8PSK / MSK / QAM16 is calculated by the following method: Calculate the bandwidth B of the signal to be identified x(t) W , x(t) represents the signal amplitude at time t. Since the symbol rate is closely related to the bandwidth, the bandwidth is used here to roughly estimate the symbol period T=1 / B W , delay the signal x(t) by half a symbol period, τ=T / 2, τ is the delay time, and get the delayed signal x(t-τ). Multiply the delayed signal and the original signal to get the delayed sequence y(t). Perform periodic spectral line detection on the delayed sequence (such as Figure 5 As shown in Figure 2), the average period of the delay sequence is calculated based on the results of periodic spectrum line detection, and the symbol rate is calculated based on the average period of the delay sequence and formula (7):

[0114] R b =T'*f s / n(7);

[0115] Among them, R b is the code rate, and T' is the average period of the delay sequence.

[0116] The main parameters and functions of the present invention are as follows:

[0117] (1) Identification signal-to-noise ratio: >=10dB;

[0118] (2) Identify modulation types: AM, FM, BPSK, QPSK, OQPSK, 8PSK, FSK, MSK, QAM16;

[0119] (3) Bit rate range: 1kHz~5MHz;

[0120] (4) Recognition accuracy: >= 90%;

[0121] (5) Parameter estimation capability: modulation frequency and modulation depth of AM; modulation frequency and frequency deviation of FM; frequency deviation and code rate of FSK; code rate of BPSK, QPSK, OQPSK, 8PSK, MSK, and QAM16.

[0122] Example 3.

[0123] The present invention provides a process for blindly identifying communication signals and estimating parameters using Example 2, including the following steps 1 to 9.

[0124] Step 1: When the system control software performs real-time frame reception of a signal in a certain frequency band, when it is necessary to identify a signal at a certain frequency point, a control instruction including the frequency point and sampling rate of the signal is issued, and the control instruction is issued to the computer processing unit via Ethernet.

[0125] Step 2 is the operation performed by the computer processing unit in Example 2, including: calculating the control parameters of the FPGA digital signal processing module according to the control instructions, and sending the control parameters to the FPGA digital signal processing module through the SRIO protocol.

[0126] Step three is all the operations performed by the FPGA digital signal processing module in Example 2.

[0127] Step 4 is all the operations performed by the frequency synthesizer unit in Example 2.

[0128] Step five is all the operations performed by the A / D acquisition module in Example 2.

[0129] Step six is ​​the operation performed by the computer processing unit in Example 2, including: the data of the low-frequency digital signal is transmitted back to the computer processing unit via the SRIO protocol, and after the computer processing unit receives the data, it forwards the data to the system control software via Ethernet.

[0130] Step seven is the operation performed by the identification module of the system control software in Example 2, including the extraction of multi-dimensional statistical features of the low-frequency digital signal. For the specific process, see Example 2.

[0131] Step eight is the operation performed by the identification module of the system control software in Example 2, including identification of the modulation type of the signal to be identified and calculation of various parameters. For the identification method and calculation method, see Example 2.

[0132] Step 9: Finally, the modulation type recognition result and parameter calculation result are sent to the display control interface. The above steps complete the blind recognition and parameter estimation process of the RF signal.

[0133] The above is only a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the technical principles of the present invention. These improvements and modifications should also be regarded as the scope of protection of the present invention.

Claims

1. A communication signal blind recognition and parameter estimation system, characterized in that: It includes receiver front end, microwave down-conversion unit, frequency synthesis unit, intermediate frequency signal processing unit, computer processing unit and system control software; The receiver front end is used to: receive the RF analog signal and input it into the microwave down-conversion unit; The frequency synthesizer unit is used to: generate a local oscillator signal and input it into the microwave down-conversion unit; The microwave down-conversion unit is used to: pre-process the RF analog signal, mix it with the local oscillator signal, obtain the intermediate frequency analog signal after low-pass filtering, and input it into the intermediate frequency signal processing unit; The intermediate frequency signal processing unit is used to: convert the intermediate frequency analog signal into an intermediate frequency digital signal, perform frequency and rate processing on the intermediate frequency digital signal according to the control parameters sent by the computer processing unit, obtain a low frequency digital signal and send it to the computer processing unit; The computer processing unit is used to: calculate the control parameters according to the control instructions issued by the system control software and send them to the intermediate frequency signal processing unit, and send the low frequency digital signal to the system control software; The system control software is used to generate and issue control instructions, perform blind identification and parameter estimation on low-frequency digital signals, and obtain modulation type and parameter estimation results.

2. The communication signal blind recognition and parameter estimation system according to claim 1, characterized in that: The preprocessing performed by the microwave down-conversion unit on the radio frequency analog signal includes limiting amplification and automatic gain control.

3. The communication signal blind recognition and parameter estimation system according to claim 1, characterized in that: The intermediate frequency signal processing unit includes an A / D sampling module and an FPGA digital signal processing module; The A / D sampling module is used to: sample the intermediate frequency analog signal to obtain an intermediate frequency digital signal, and input the intermediate frequency digital signal into the FPGA digital signal processing module; The FPGA digital signal processing module is used to: receive the control parameters, including the frequency word and the decimation multiple, down-convert and down-sample the intermediate frequency digital signal according to the control parameters to obtain a low-frequency digital signal, and send the low-frequency digital signal to the computer processing unit.

4. The communication signal blind recognition and parameter estimation system according to claim 3, characterized in that: The intermediate frequency digital signal is high-rate multi-channel data, and the low-frequency digital signal is low-rate single-channel data.

5. The communication signal blind recognition and parameter estimation system according to claim 3, characterized in that: The FPGA digital signal processing module and the computer processing unit communicate with each other using the SRIO protocol, and the computer processing unit and the system control software communicate with each other using Ethernet.

6. The communication signal blind recognition and parameter estimation system according to claim 1, characterized in that: The system control software includes display control software and identification module; The display and control software is used to: issue control instructions according to the frequency band and bandwidth of the radio frequency analog signal, the control instructions include frequency and sampling rate, receive the modulation type and parameter estimation results output by the recognition module and display them on the interface; The identification module is used to: obtain a signal to be identified by preprocessing the low-frequency digital signal in the frequency domain, extract multidimensional statistical features of the signal to be identified, input the multidimensional statistical features into a pre-trained modulation type identification model for prediction to obtain the modulation type, and estimate the modulation parameters of the signal to be identified according to the predicted modulation type to obtain a parameter estimation result.

7. The communication signal blind recognition and parameter estimation system according to claim 6, characterized in that: The frequency domain preprocessing includes removing the DC component and accurately estimating and down-converting the carrier frequency.

8. The communication signal blind recognition and parameter estimation system according to claim 6, characterized in that: The multidimensional statistical features include time domain features, amplitude features, phase features and frequency domain features; The time domain characteristics include fourth-order cumulative quantities and sixth-order cumulative quantities; The amplitude characteristics include the compactness of the zero-centered normalized instantaneous amplitude, the standard deviation of the absolute value of the zero-centered normalized instantaneous amplitude, and the standard deviation of the zero-centered normalized non-weak signal segment instantaneous amplitude; The phase feature includes the standard deviation of the zero-center normalized non-weak signal segment instantaneous phase nonlinear component and the standard deviation of the absolute value of the zero-center normalized non-weak signal segment instantaneous phase nonlinear component; The frequency domain features include the zero-centered normalized instantaneous frequency compactness, the standard deviation of the absolute value of the zero-centered normalized non-weak signal instantaneous frequency, the proportion of the top points of the power spectrum, and the standard deviation of the top point set of the power spectrum.

9. The communication signal blind recognition and parameter estimation system according to claim 6, characterized in that: The modulation type recognition model is a random forest model; The training of the modulation type recognition model includes: Constructing a training set based on the multidimensional statistical features extracted from the historically acquired low-frequency digital signals after the frequency domain preprocessing; The training set is input into the modulation type recognition model, and the trained modulation type recognition model is obtained through iterative training.

10. The communication signal blind recognition and parameter estimation system according to claim 6, characterized in that: The estimating the modulation parameters of the signal to be identified according to the predicted modulation type to obtain the parameter estimation result includes: If the predicted modulation type is AM, the signal to be identified is an AM signal, and the AM carrier frequency is calculated by the following method: the spectrum of the AM signal is calculated, the threshold of the AM carrier frequency is set according to the amplitude of the spectrum, the carrier frequencies at both ends of the AM signal spectrum are obtained, and the symmetry of the AM carrier frequency is used to calculate the average distance from the carrier frequencies at both ends of the AM signal spectrum to the center frequency point, and the AM carrier frequency is calculated by formula (1): F AM =D* f s / n(1); Among them, f AM is the AM carrier frequency, D is the average distance from the carrier frequency at both ends of the AM signal spectrum to the center frequency, and f s is the sampling rate, n is the number of signal points; The AM depth is calculated by the following method: a middle section of the AM signal is selected to calculate the modulus of the complex signal, and a ratio of the difference between the maximum values ​​of the modulus and the sum of the maximum values ​​is calculated, where the ratio is the AM depth; If the predicted modulation type is FM, the signal to be identified is an FM signal, and the FM frequency deviation is calculated by the following method: the spectrum of the FM signal is calculated, and the threshold of the FM frequency deviation is set according to the amplitude of the spectrum, and the frequencies at the two ends of the FM spectrum above the threshold are obtained. The symmetry of the FM spectrum is used to calculate the average distance from the two ends of the FM signal spectrum to the center frequency point, and the FM frequency deviation is calculated by formula (2): F FM =D1* f s / n(2); Among them, F FM is the FM frequency deviation, D1 is the average distance from the frequencies at the two ends of the FM signal spectrum to the center frequency; The FM carrier frequency is calculated by the following method: the center frequency of the FM signal is accurately estimated to obtain the accurately estimated carrier frequency, the FM signal is down-converted using the accurately estimated carrier frequency, and its instantaneous phase sequence is calculated modulo 2π, the value range of the instantaneous phase sequence is (-π, π), and the instantaneous phase sequence is a folded phase sequence. The non-folded phase sequence is recovered from the instantaneous phase sequence according to the corrected phase sequence, and then the instantaneous frequency sequence is estimated according to the non-folded phase sequence. Finally, the spectrum of the instantaneous frequency sequence is calculated, and the frequency point corresponding to the maximum value of the spectrum amplitude of the instantaneous frequency sequence is the FM carrier frequency; The modified phase sequence is calculated by formula (3): （3）； Where C(i) is the corrected phase sequence corresponding to the i-th signal point (i=1,…,n), and θ(i) is the instantaneous phase sequence corresponding to the i-th signal point; The method of recovering the non-folding phase sequence from the instantaneous phase sequence according to the modified phase sequence is performed by formula (4): Ф(i)=θ(i)+ C(i-1)(4) Where Ф(i) is the unfolded phase sequence corresponding to the i-th signal point; The instantaneous frequency sequence of the unfolded phase sequence is calculated by formula (5): （5）； Among them, f(i) is the instantaneous frequency sequence corresponding to the i-th signal point, T s is the sampling period, T s =1 / f s ; If the predicted modulation type is FSK, the signal to be identified is an FSK signal. The FSK frequency deviation is calculated by the following method: the spectrum of the FSK signal is calculated, and the threshold of the FSK frequency deviation is set according to the amplitude of the spectrum. The frequencies at the two ends of the FSK signal spectrum above the threshold are obtained. The symmetry of the FSK spectrum is used to calculate the average distance from the impact at both ends of the FSK signal spectrum to the center frequency point. The FSK frequency deviation is calculated by formula (6): F FSK =D2* f s / n(6); Among them, f FSK is the FSK frequency deviation, and D2 is the average distance from the impulses at both ends of the FSK signal spectrum to the center frequency point; The symbol rate of FSK / BPSK / QPSK / OQPSK / 8PSK / MSK / QAM16 is calculated by the following method: Calculate the bandwidth of the signal to be identified, according to the formula T=1 / B W Calculate the symbol period, B W is the bandwidth. The signal to be identified is delayed by half a symbol period and then multiplied with the original signal to be identified to obtain a delayed multiplication sequence. The delayed multiplication sequence is subjected to periodic spectrum line detection, and the average period of the delayed multiplication sequence is calculated based on the result of periodic spectrum line detection. Based on the average period of the delayed multiplication sequence, the symbol rate is calculated using formula (7): R b =T’* f s / n(7); Among them, R b is the symbol rate, and T' is the average period of the delay sequence.

Images