Signal transmission protocol intelligent identification method and device for radio monitoring

Abstract

The invention discloses a signal transmission protocol intelligent identification method and device for radio monitoring, relates to the field of radio detection, and is used for improving the robustness, universality and practicability of signal protocol identification. After normalization processing is carried out on signal power, discrete and continuous initial classification is carried out based on time domain features of signals, fine classification is carried out on discrete types and then based on high-power spectrum features and time-frequency features of the signals, and fine classification is carried out on continuous types and then based on the time-frequency features, the high-power spectrum features and self-correlation features of the signals. According to the method, the robustness, the universality and the practicability of signal protocol identification are improved.

Description

Technical Field

[0001] The invention relates to the field of radio detection, and in particular to a signal transmission protocol intelligent identification method and device for radio monitoring. Background Art

[0002] With the rapid development of wireless communication technology, the number of wireless access devices has also increased significantly. Generally speaking, the communication between wireless access devices will follow the specific frequency range and modulation method designed by each protocol. If a signal of a different protocol type suddenly appears, such as a change in frequency band or modulation method, it may indicate abnormal activities in the scene, such as unauthorized access or malicious attacks and interference. Therefore, the identification of the received signal protocol is crucial to ensure the security of wireless communications. With the rapid development of wireless communication protocols, how to efficiently monitor large-scale protocol signals has become a greater challenge.

[0003] Currently known signal protocol identification methods mostly focus on protocol identification at the application layer, usually by detecting some parameters of the signal for feature analysis. For example, a communication protocol identification method based on deep learning is proposed in the Chinese patent document with publication number CN108712397A, which extracts the features of the signal I / Q sequence and feeds it into the convolutional recognition network to identify the signal communication protocol type. This type of method does not take into account the changes in signal parameters caused by channel changes, which will cause the protocol to remain unchanged while the signal parameters will change dramatically, resulting in a deviation between the recognition result and the actual result; in addition, the relevant parameters or modulation methods of some protocols are the same, which will also make this type of recognition method unable to accurately identify the signal protocol. Therefore, this type of recognition method has poor generalization.

[0004] In addition, there are also studies showing that the signal protocol can be identified through the time-frequency characteristics of the signal. For example, a method for identifying IoT communication protocols based on the combination of time and frequency domains is proposed in the Chinese patent document with publication number CN118646687A. It realizes the identification of IoT communication protocols by combining the time-frequency characteristics of the protocol signal. However, the samples used in such methods are almost all from simulated QAM (Quadrature Amplitude Modulation) data, and the real protocol signals from different signal sources in complex aerial scenes and the possible influence of various noises and interferences on the algorithm are not considered, which makes such methods less practical.

[0005] In addition, there is another type of method that uses various parameters of protocol signals, including modulation mode, frequency and other parameters, to establish protocol feature vectors, calculates entropy to quantize the values ​​of each parameter, and uses SVM to classify and identify signal protocols. For example, a Chinese patent document with publication number CN117596193A proposes a message protocol identification method and system based on a neural network, which screens and analyzes message sequence features to identify the communication protocol of the data packet. However, this type of method is highly dependent on the recognition accuracy of each signal parameter. The presence of various types of interference in complex aerial scenes will lead to error accumulation, which seriously reduces the accuracy of protocol recognition and has poor robustness. Summary of the invention

[0006] The object of the present invention is to provide a method and device for intelligently identifying signal transmission protocols for radio monitoring in order to improve at least one of the robustness, versatility and practicality of signal protocol identification in order to address all or part of the above-mentioned problems.

[0007] The technical solution adopted by the present invention is as follows:

[0008] A signal transmission protocol intelligent identification method for radio monitoring, comprising:

[0009] S1. Perform power normalization on the received signal S(n) to obtain a unit power signal S a (m);

[0010] S2, based on S a The time domain characteristics of (n) determine whether S(n) is a time discrete signal or a time continuous signal;

[0011] S3, for time-discrete signals, based on S a The high-order spectrum characteristics of S(n) determine whether S(n) is {GSM}, {802.11b}, {CDMA2000} or {NR, TD-LTE, 802.11a / g / n}; if S(n) belongs to {NR, TD-LTE, 802.11a / g / n}, then based on S a The time-frequency characteristics of (n) determine whether S(n) is {NR}, {TD-LTE} or {802.11a / g / n};

[0012] S4. For time-continuous signals, based on S a The time-frequency characteristics of S(n) determine whether S(n) is a frequency-hopping signal or a non-frequency-hopping signal; if S(n) is a frequency-hopping signal, then based on S a The high-order spectrum characteristics of S(n) determine whether S(n) is {BR / EDR} or {BLE}; if S(n) is a non-frequency hopping signal, then based on S aThe autocorrelation characteristics of S(n) determine whether S(n) is {FD-LTE} or {Noise}.

[0013] In order to solve all or part of the above problems, the present invention also provides a device, including a processor and a storage medium, wherein the storage medium stores computer instructions, and when the processor runs the computer instructions, it can execute the above-mentioned signal transmission protocol intelligent identification method for radio monitoring.

[0014] In summary, due to the adoption of the above technical solution, the beneficial effects of the present invention are:

[0015] The present invention identifies the protocol type of the received signal in multiple levels in sequence according to the characteristics of different signal protocols, which improves the versatility of the identification method compared to the method of directly extracting signal parameters (such as I / Q sequences) for identification. In addition, the present invention fully considers the multimodal characteristics of the signal, deeply combines the signal protocol format to identify the signal protocol type, and improves the robustness of the identification method. Moreover, the present invention is directly applied to the real air interface signals collected by the receiver, taking into account the impact of the actual air interface signals from different transmitting devices in complex aerial scenes, and verifies the practicality of the identification method. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] The present invention will now be described by way of example with reference to the accompanying drawings, in which:

[0017] Figure 1 It is a flow chart of the intelligent identification method of signal transmission protocol for radio monitoring according to an embodiment of the present application.

[0018] Figure 2 It is a signal flow diagram of one embodiment of the signal transmission protocol intelligent identification method for radio monitoring provided by the present application.

[0019] Figure 3 It is a flowchart of the time domain identification process provided by an embodiment of the present application.

[0020] Figure 4 It is a structural diagram of the MobileNet image recognition network provided in an embodiment of the present application.

[0021] Figure 5 It is a waveform diagram of 10 signals including GSM and CDMA2000 after time domain modulo values ​​are taken in the embodiment of the present application.

[0022] Figure 6 It is a time domain I / Q waveform and square spectrum characteristic diagram of the GSM signal in the embodiment of the present application.

[0023] Figure 7 It is a square spectrum characteristic diagram corresponding to four different transmission rates of the 802.11b protocol in the embodiment of the present application.

[0024] Figure 8 It is a time domain I / Q waveform and fourth power spectrum characteristic diagram of the CDMA2000 signal in the embodiment of the present application.

[0025] Fig. 9 It is a time-frequency characteristic diagram of NR, TD-LTE, and 802.11a / g / n in the embodiment of the present application.

[0026] Fig.10 It is a time-frequency characteristic diagram of the Bluetooth signal and a square spectrum characteristic diagram of the BLE signal and the BR / EDR signal in the embodiment of the present application.

[0027] Fig.11 It is an autocorrelation characteristic diagram of the autocorrelation operation performed on FD-LTE and pure noise signal in the embodiment of the present application.

[0028] Fig.12 This is a confusion matrix diagram for identifying 9 common protocol signals and air interface noise in an embodiment of the present application. DETAILED DESCRIPTION

[0029] All features disclosed in this specification, or steps in all methods or processes disclosed, except mutually exclusive features and / or steps, can be combined in any manner.

[0030] Any feature disclosed in this specification (including any additional claims and abstract), unless otherwise stated, may be replaced by other equivalent or alternative features having similar purposes. That is, unless otherwise stated, each feature is only an example of a series of equivalent or similar features.

[0031] In response to the problem that existing signal protocol identification methods perform poorly in terms of versatility, robustness and practicality, the embodiments of the present application provide a method and device for intelligently identifying signal transmission protocols for radio monitoring, so as to improve the performance of the signal protocol identification method in one or more aspects of versatility, robustness and practicality.

[0032] In some embodiments, Figure 1 As shown, the signal transmission protocol intelligent identification method for radio monitoring provided by the embodiment of the present application includes the following steps:

[0033] S1. Perform power normalization processing on the received signal S(n) (n represents the sampling point) to obtain a unit power signal S a (n).

[0034] The received signal S(n) is a signal for which protocol type identification is required. The received signal is a protocol signal in a real non-cooperative aerial scenario, and contains pure noise and interference type signals.

[0035] As an optional implementation, the received signal S(n) is normalized using the following formula:

[0036]

[0037] Wherein, L is the number of sampling points of S(n), that is, the length of the received signal S(n).

[0038] S2, based on S a The time domain characteristics of S(n) determine whether S(n) is a time-discrete signal or a time-continuous signal.

[0039] Since some protocols use TDMA or the frame structure has a clear time interval, this type of signal must be discontinuous in time, which is specifically manifested as the change of signal amplitude, while other signals that do not involve this feature are still continuous in time. Therefore, the initial classification results of S(n) in time discrete signals and time continuous signals can be obtained based on the time domain characteristics of the signal.

[0040] As an optional implementation, Figure 2 As shown, step S2 includes:

[0041] S2.1. Obtain S a (n), and sequentially divide the I / Q sequence into multiple I / Q sequence segments.

[0042] The so-called I / Q sequence is the signal S A (n) Characteristic components in the in-phase and quadrature directions. In step S2.1, the I / Q sequence is sequentially divided into a plurality of I / Q sequence segments according to a set length, and the last I / Q sequence segment is reserved for use.

[0043] S2.2. Send the I / Q sequence to the stacked dilated convolution module and the first fully connected module for processing to obtain the first feature.

[0044] In some possible implementations, such as Figure 3 As shown, the above-mentioned stacked dilated convolution module sequentially includes a one-dimensional convolution layer Conv1d, an activation function GELU, a one-dimensional maximum pooling layer MaxPool1d, a one-dimensional convolution layer Conv1d, an activation function GELU, a one-dimensional maximum pooling layer MaxPool1d, a one-dimensional convolution layer Conv1d, an activation function GELU, a one-dimensional maximum pooling layer MaxPool1d, a one-dimensional convolution layer Conv1d, an activation function GELU, a one-dimensional maximum pooling layer MaxPool1d, a one-dimensional convolution layer Conv1d, an activation function GELU, and a one-dimensional maximum pooling layer MaxPool1d.

[0045] The above-mentioned first fully connected layer includes a linear layer Linear, an activation function ReLU, and a linear layer Linear in sequence.

[0046] S2.3. Send the last I / Q sequence segment to the timing feature convolution module for processing to obtain the second feature.

[0047] In some possible implementations, such as Figure 3 As shown, the above-mentioned time series feature convolution module includes a one-dimensional convolution layer Conv1d, a one-dimensional average pooling layer AvgPool1d, a KAN network layer KANLayer, a one-dimensional batch normalization layer BatchNorm1d, an activation function ReLU, a one-dimensional convolution layer Conv1d, a one-dimensional average pooling layer AvgPool1d, a KAN network layer KANLayer, an LSTM network layer, an activation function ReLU, a one-dimensional batch normalization layer BatchNorm1d, a KAN network layer KANLayer, and an activation function ReLU.

[0048] The above KAN network layer is a neural network architecture, which is the application of Kolmogorov-Arnold representation theorem in neural networks. Kolmogorov-Arnold representation proves that a multivariate continuous function can be written as a nested superposition of single variable continuous functions. Specifically:

[0049]

[0050] In the formula, Φ q is called the external function, φ p,q is called the internal function, both of which are continuous single-variable functions. The relevant explanation of the above formula is common knowledge and will not be introduced in detail here. The above formula shows that in the real number domain, only the sum is the only true multivariate function, and all other functions can be represented by the sum of single-variable functions. Therefore, the KAN network has a strong fitting ability. When processing time series analysis, the KAN network has higher accuracy and interpretability in processing complex function fitting, solving partial differential equations, etc. by learning complex relationships and nonlinear patterns in the data.

[0051] In addition, the above-mentioned LSTM (Long Short-Term Memory) layer is used for the analysis of time series data, which can effectively capture the temporal dependencies in the data. It includes a forget gate, an input gate, and an output gate. The relevant descriptions are also common knowledge.

[0052] like Figure 3 As shown, the above-mentioned stacked dilated convolution module, the first fully connected module, the time series feature convolution module and the second fully connected module together constitute a time domain signal recognition network based on LSTM (Long-Short Term Memory Network).

[0053] S2.4. The first feature and the second feature are integrated and sent to the second fully connected module to obtain the classification results of S(n) in time-discrete signals and time-continuous signals.

[0054] In some feasible implementations, the first feature and the second feature are concatenated and fused, and then sent to a second fully connected module for classification. The structure of the second fully connected module is the same as that of the first fully connected module.

[0055] After the above classification, the S a The time domain characteristics of (n) preliminarily classify the received signal S(n) into a time discrete signal or a time continuous signal, narrowing the identification range of the signal protocol type. In an embodiment of the present application, the possible protocol types of the received signal include {GSM, 802.11a / g / n, 802.11b, CDMA2000, TD-LTE, NR, BR / EDR, BLE, FD-LTE, Noise}. After the above preliminary classification, the received signal S(n) either belongs to a time discrete signal {GSM, 802.11a / g / n, 802.11b, CDMA2000, TD-LTE, NR}, or a time continuous signal {BR / EDR, BLE, FD-LTE, Noise}.

[0056] S3, for time-discrete signals, based on S a The high-order spectrum characteristics of S(n) determine whether S(n) is {GSM}, {802.11b}, {CDMA2000} or {NR, TD-LTE, 802.11a / g / n}; if it belongs to {NR, TD-LTE, 802.11a / g / n}, then based on S a The time-frequency characteristics of (m) determine whether S(n) is {NR}, {TD-LTE} or {802.11a / g / n}.

[0057] To facilitate the distinction in the following description, S a1 (n) represents the time-discrete signal belonging to {GSM, 802.11a / g / n, 802.11b, CDMA2000, TD-LTE, NR}, with S a2 (n) represents the time-discrete signal belonging to {802.11a / g / n, CDMA2000, TD-LTE, NR}, with S a3 (n) represents a time-discrete signal belonging to {802.11a / g / n, TD-LTE, NR}.

[0058] In the embodiment of the present application, for a time-discrete signal, based on S a1The high-order spectrum characteristics of (n) determine whether S(n) is {GSM}, {802.11b}, {CDMA2000} or {NR, TD-LTE, 802.11a / g / n}. As an optional implementation, the process includes:

[0059] S3.1, based on S a1 The square spectrum characteristics of (n) determine whether S(n) is {GSM}, {802.11b}, or belongs to {CDMA2000, NR, TD-LTE, 802.11a / g / n}.

[0060] In some feasible implementations, in step S3.1, signal S a1 (n) Quadratic fast Fourier transform and modulus processing to extract signal S a1 The square spectrum characteristic F of (n) a1 (k), expressed as:

[0061]

[0062] Then the square spectrum feature F a1 (k) is input to the MobileNet image recognition network, which is obtained from F a1 (k) to classify S(n). Since the GSM protocol signal uses GMSK modulation with a modulation coefficient of h = 0.5, there will be two obvious discrete spectral lines on both sides of the image in the square spectrum feature, which can be used for classification. That is, if F a1 If two discrete spectral lines appear on both sides of the square spectrum feature image of (k), the protocol type of S(n) is {GSM}. The 802.11b protocol signal uses four different modulation methods according to the rate, including DSSS direct spread and CCK keying modulation. They have an obvious discrete spectral line at the zero frequency in the square spectrum feature, which can be used for classification. Therefore, in step S3.1, the final recognition result of the MobileNet image recognition network for S(n) is {GSM}, {802.11b} or {CDMA2000, NR, TD-LTE, 802.11a / g / n}.

[0063] As an optional implementation, Figure 4 As shown in the figure, the MobileNet image recognition network consists of a deep convolution module a, a deep convolution module b, and a classification module:

[0064] The deep convolution module a includes, in sequence, a two-dimensional convolution layer Conv2d, a two-dimensional convolution layer Conv2d, a two-dimensional convolution layer Conv2d, a two-dimensional maximum pooling layer Maxpool2d, an activation function ReLU, a two-dimensional convolution layer Conv2d, a two-dimensional convolution layer Conv2d, a two-dimensional convolution layer Conv2d, a two-dimensional maximum pooling layer Maxpool2d, and an activation function ReLU.

[0065] The deep convolution module b includes a two-dimensional convolution layer Conv2d, a two-dimensional batch normalization layer BatchNorm2d, an activation function ReLU, a two-dimensional convolution layer Conv2d, a two-dimensional batch normalization layer BatchNorm2d, an activation function ReLU, a two-dimensional convolution layer Conv2d, a two-dimensional batch normalization layer BatchNorm2d, an activation function ReLU, a two-dimensional convolution layer Conv2d, a two-dimensional batch normalization layer BatchNorm2d, and an activation function ReLU.

[0066] The classification module includes a two-dimensional average pooling layer Avgpool2d, a flat layer Flatten, a linear layer Linear, an activation function ReLU, a linear layer Linear, an activation function ReLU, and a linear layer Linear.

[0067] S3.2, if S(n) belongs to {CDMA2000, NR, TD-LTE, 802.11a / g / n}, then based on S a2 The fourth power spectrum characteristics of (n) determine whether S(n) belongs to {CDMA2000} or {NR, TD-LTE, 802.11a / g / n}.

[0068] In some feasible implementations, in step S3.2, signal S a2 (n) The fourth power fast Fourier transform and modulus processing are used to extract the fourth power spectrum feature F of the signal. a2 (k), expressed as:

[0069]

[0070] Then the quartic spectrum feature F a2 (k) is input into the MobileNet image recognition network, and the MobileNet image recognition network is used to obtain the image from F a2(k) is used to classify S(n). Since the CDMA2000 forward link waveform uses QPSK modulation technology, there are no obvious features in the quadratic spectrum feature, but three discrete spectral lines will appear in the quartic spectrum feature, and there is an obvious discrete spectral line at zero frequency, so it can be used to identify CDMA2000 protocol signals. In step S3.2, the final classification result of S(n) by the MobileNet image recognition network is {CDMA2000} or {NR, TD-LTE, 802.11a / g / n}.

[0071] In addition, step S3 further includes:

[0072] S3.3. If S(n) belongs to {NR, TD-LTE, 802.11a / g / n}, then based on S a3 The time-frequency characteristics of (n) determine whether S(n) is {NR}, {TD-LTE} or {802.11a / g / n}.

[0073] As an optional implementation, by a3 (n) Perform short-time Fourier transform to obtain its time-frequency characteristics S α3 (m,f), expressed as:

[0074] Where m and f represent time and frequency respectively, g(n) is the window function, and R is a constant.

[0075] In some feasible implementations, according to S α3(m,f) Draw the time-frequency feature map, send the time-frequency feature map to the MobileNet image recognition network, and classify S(n) according to the two-dimensional information of time domain and frequency domain. 802.11a uses OFDM modulation, and in the design of its frame structure, it has a fixed time interval of 0.1ms when sending data packets continuously. Therefore, 802.11a can be identified by this feature in the time-frequency feature map of fixed time length; 802.11g uses the same modulation method as 802.11a, the difference is that it is located in the 2.4G frequency band, while 802.11a is located in the 5G frequency band; 802.11n is OFDM modulation with higher speed and greater throughput, so these three protocol types can be grouped together for identification. For TD-LTE, the length of a wireless frame in its frame structure is 10ms, which contains 10 subframes in total. Each subframe contains 2 time slots, and there are three types of uplink subframes, downlink subframes and special subframes. LTE has 7 fixed subframe configuration modes in time division multiplexing, including DSUUU DSUUU, DSUUD DSUUD, DSUUU DDDDD, etc. In other words, in the time dimension, TD-LTE has a flexible transmission interval, which is significantly different from the 802.11 series protocol cluster, and the length of each uplink subframe is 1ms, which is also much longer than the time interval designed by the 802.11 protocol. The 802.11a / g / n and TD-LTE protocol types can be distinguished based on this time domain feature. Finally, the physical resources designed by the NR protocol include the SSB synchronization block necessary for signal transmission, which consists of four parts: the primary synchronization signal PSS, the secondary synchronization signal SSS, the demodulation reference signal DMRS, and the physical broadcast channel PBCH. It occupies a total of 4 OFDM symbols in the time domain and 240 subcarriers in the frequency domain. Its particularity lies in the PSS, which is located in the 56th to 182nd subcarrier part of the 0th OFDM symbol on the time-frequency resource, and the remaining subcarrier positions are all empty, while the 240 subcarriers on the 1st to 3rd OFDM symbols are generally fully occupied. Therefore, the SSB synchronization block can be found on the time-frequency feature diagram based on the frequency domain features to determine the type of NR protocol signal. In step S3.3, the final recognition result of the MobileNet image recognition network for S(n) is {NR}, {TD-LTE}, {802.11a / g / n}.

[0076] S4. For time-continuous signals, based on S a The time-frequency characteristics of S(n) determine whether S(n) is a frequency-hopping signal or a non-frequency-hopping signal; if S(n) is a frequency-hopping signal, then based on S a The high-order spectrum characteristics of S(n) determine whether S(n) is {BR / EDR} or {BLE}; if S(n) is a non-frequency hopping signal, then based on S aThe autocorrelation characteristics of S(n) determine whether S(n) is {FD-LTE} or {Noise}.

[0077] For the convenience of distinguishing in the following description, S a4 (n) represents the time-continuous signal belonging to {BR / EDR, BLE, FD-LTE, Noise}, with S a5 (n) represents a time-continuous signal belonging to a frequency-hopping signal, with S a6 (n) represents a time-continuous signal which is a non-frequency hopping signal.

[0078] Step S4 is subdivided into three processes:

[0079] S4.1. For time-continuous signals, based on S a4 The time-frequency characteristics of S(n) determine whether S(n) is a frequency-hopping signal or a non-frequency-hopping signal.

[0080] As an optional implementation, the signal S a4 (n) Perform short-time Fourier transform to obtain its time-frequency characteristics S α4 (m,f), expressed as:

[0081]

[0082] In some feasible implementations, according to S α4 (m,f) Draw the time-frequency feature graph, send the time-frequency feature graph to the MobileNet image recognition network, classify S(n) according to the two-dimensional information of time domain and frequency domain, and identify whether S(n) is a frequency hopping signal or a non-frequency hopping signal. {BR / EDR, BLE} are both Bluetooth protocols. For the Bluetooth protocol, frequency hopping is used in the transmission frequency band. The main frequency components change over time, while non-frequency hopping signals do not change over time. Therefore, frequency hopping signals and non-frequency hopping signals can be intuitively classified from the time-frequency feature graph. That is, in step S4.1, the final recognition result of S(n) by the MobileNet image recognition network is {BR / EDR, BLE} or {FD-LTE, Noise}.

[0083] S4.2. If S(n) is a frequency hopping signal, then based on S a5 The high-order spectral characteristics of (n) determine whether S(n) is {BR / EDR} or {BLE}.

[0084] As an optional implementation, in step S4.2, based on S a5 The squared spectral characteristics of S(n) determine whether S(n) is {BR / EDR} or {BLE}.

[0085] In some possible implementations, the signal Sa5 (n) Quadratic fast Fourier transform and modulus processing to extract signal S a5 The square spectrum characteristic F of (n) a5 (k), expressed as:

[0086]

[0087] Then the square spectrum feature F a5 (k) is input to the MobileNet image recognition network, which is obtained from F a5 (k) is used to classify S(n). The Bluetooth protocol includes the classic Bluetooth waveform BR / EDR and the low-power Bluetooth waveform BLE. Both use GFSK modulation, but the BLE signal is similar to the GSM modulation method, and the modulation coefficient is also h=0.5. Therefore, in the square spectrum feature, the BLE signal also has two obvious discrete spectral lines on both sides of the image, but the modulation coefficient of the classic Bluetooth BR / EDR waveform is between 0.28 and 0.35, and there is no such obvious discrete spectral line. This feature can be fully utilized to classify the two. In step S4.2, the final recognition result of the MobileNet image recognition network for S(n) is {BR / EDR}, {BLE}.

[0088] S4.3. If S(n) is a non-frequency hopping signal, then based on S a6 The autocorrelation characteristics of S(n) determine whether S(n) is {FD-LTE} or {Noise}.

[0089] For non-frequency hopping signals, only two protocol types are left: {FD-LTE} and {Noise}. As an optional implementation, for signal S a6 (n) Perform autocorrelation operation to obtain S a6 (n) autocorrelation characteristic R s (τ), expressed as:

[0090]

[0091] Where τ is the delay time, N corr The time limit.

[0092] Since the FD-LTE signal uses OFDM modulation, in the OFDM system, a guard interval is usually added to reduce the interference between symbols. In practical applications, the cyclic prefix method is generally used. The data segment at the end of the OFDM symbol is copied to the guard interval at the transmitter to form a cyclic prefix CP, thereby ensuring that each subcarrier can maintain a complete cycle within the FFT time window when the CP is removed and FFT is processed at the receiver. This copied guard interval can be detected using autocorrelation features, and compared with the irregular pure noise signal {Noise}, the two can be classified based on this feature. In step S4.3, the final recognition result of the MobileNet image recognition network for S(n) is {FD-LTE}, {Noise}.

[0093] It should be noted that the LSTM-based signal recognition network and the MobileNet image recognition network mentioned in the embodiments of the present application are both trained using data samples before use. The training method is a well-known technology in the art, so it is not described in detail in the embodiments of the present application.

[0094] In addition, an embodiment of the present application also provides a device, which includes a processor and a storage medium, in which computer instructions are stored. When the processor runs the computer instructions, it can execute the signal transmission protocol intelligent identification method for radio monitoring of the above embodiment.

[0095] The embodiment of the present application also experimentally verifies the proposed method. In this experiment, the performance of the algorithm is experimentally verified around 10 common air protocol signal types, including 2G mobile communication standard (GSM), 3G mobile communication system (CDMA2000), 4G mobile communication system (time division duplex system: TD-LTE, frequency division duplex system: FD-LTE), 5G mobile communication system (NR), classic Bluetooth waveform (BR / EDR), low-power Bluetooth waveform (BLE), wireless LAN protocol cluster (802.11a, 802.11b, 802.11g, 802.11n), and pure noise signal (Noise). The signal data sets used in the embodiments of the present application are all derived from actual complex air communication scenarios, collected by the Siyi 3943B spectrum monitoring receiver with antenna, and the signals come from various different sending devices, which fully considers the practicality of this algorithm. Among them, when collecting signals, the Siyi 3943B spectrum monitoring receiver needs to select a suitable demodulation bandwidth according to the bandwidth of different protocol signals, and there is a certain mapping relationship between the demodulation bandwidth and the sampling rate, as shown in Table 1.

[0096] Table 1 Correspondence between demodulation bandwidth and sampling rate

[0097] Demodulation bandwidth Sampling rate 300KHz 1.4MHz 500KHz 2MHz 800KHz 3.5MHz 1000KHz 4MHz 1250KHz 5.6MHz 1500KHz 7MHz 2000KHz 7MHz 5000KHz 14MHz 8000KHz 28MHz 10000KHz 28MHz 12500KHZ 28MHz 15000KHZ 28MHz 20000KHZ 28MHz

[0098] When constructing the data set of all protocol signals, the different bandwidths of each protocol in the actual scenario are taken into consideration. Therefore, each time an air signal is received, in order to avoid mutual interference between different protocol signals, it is necessary to receive only one protocol within the demodulation bandwidth, and different protocol signals come from different frequency bands. The parameters are shown in Table 2.

[0099] Table 2 Frequency band parameters for receiving air protocol signals

[0100]

[0101]

[0102] According to the specific implementation of the embodiment of the present application, the present application can identify a total of 9 types of protocol signals and 1 type of empty noise. Among them, WIFI-802.11a / g / n is classified as one, WIFI-802.11b is classified as one, classic Bluetooth BR / EDR and low-power Bluetooth BLE are classified as one, TD-LTE standard and FD-LTE standard are classified as one type for identification. The number of signal sampling points L used in this application to identify the protocol signal type is 16384.

[0103] In the experimental verification of the LSTM-based time domain signal recognition network, the signal S a (n) The number of sampling points L = 16384, the sequence dimension fed into the stacked dilated convolution module is 2 × 16384, and the sequence dimension fed into the time series feature convolution module is 2 × 512. The network parameters of the time domain signal recognition based on LSTM are shown in Table 3, where the linear layer output dimension of the classification module is l = 2, representing the probability that the recognition result may be {GSM, 802.11a / g / n, 802.11b, CDMA2000, TD-LTE, NR} and {BR / EDR, BLE, FD-LTE, Noise}. In addition, in order to distinguish the layers of the same structure in the same module, a symbol is added after the layer for distinction.

[0104] Table 3 Parameters of LSTM-based time domain signal recognition network

[0105]

[0106]

[0107]

[0108] In the experimental verification of the MobileNet image recognition network, the signal S is extracted a1 (n), S a5 (n) quadratic spectrum, extract signal S a2(n) the fourth power spectrum, extracting the signal S a3 (n), S a4 (n) time-frequency characteristics, extract signal S a6 The number of signal sampling points used when calculating the autocorrelation feature of (n) is L = 16384, and the image dimension of the input network is 64×64×3. The network parameters of the MobileNet image recognition network are shown in Table 4, where the final output dimension of the linear layer of the classification module is l = L, which varies according to different input requirements. a1 (n), the final output dimension l = 3, representing the probability that the recognition result may be {GSM}, {802.11b}, {CDMA2000, NR, TD-LTE, 802.11a / g / n}; for signal S a2 (n), the final output dimension l = 2, representing the probability that the recognition result may be {CDMA2000}, {NR, TD-LTE, 802.11a / g / n}; for signal S a3 (n), the final output dimension l = 3, representing the probability that the recognition result may be {NR}, {TD-LTE}, {802.11a / g / n}; for signal S a4 (n), the final output dimension l = 2, representing the probability that the recognition result may be {BR / EDR, BLE} and {FD-LTE, Noise}; for signal S a5 (n), the final output dimension l = 2, representing the probability that the recognition result may be {BLE} and {BR / EDR}; for signal S a6 (n), the final output dimension l = 2, representing the probability that the recognition result may be {FD-LTE} and {Noise}.

[0109] Table 4 MobileNet image recognition network parameters

[0110]

[0111]

[0112]

[0113] Experiment 1: Figure 5 It can be seen that the waveforms after time domain I / Q modulo sampling of GSM, CDMA2000, TD-LTE, NR, 802.11a / g / n, 802.11b, BR / EDR, BLE, FD-LTE, and Noise signals can be clearly divided into two types of signals: time continuous and time discontinuous.

[0114] Experiment 2: Figure 6It can be seen that this is the time domain I / Q waveform and square spectrum of the GSM signal, and the two discrete spectral lines on both sides of the spectrum can be clearly seen.

[0115] Experiment 3: Figure 7 It can be seen that these are the square spectrum images of the protocol signals corresponding to the four different transmission rates of the 802.11b protocol. It can be seen that there is always a discrete spectrum line at the zero frequency.

[0116] Experiment 4: Figure 8 It can be seen that this is the time domain I / Q waveform and the fourth power spectrum of the CDMA2000 signal. It can be clearly seen that there is a discrete spectral line at zero frequency in the middle of the spectrum.

[0117] Experiment 5: Fig. 9 It can be seen that the time-frequency diagrams of NR, TD-LTE, and 802.11a / g / n, respectively, clearly show the position of the SSB synchronization block in the physical resource of the NR signal, as well as the different time interval lengths of TD-LTE and 802.11a / g / n.

[0118] Experiment 6: Fig.10 It can be seen that these are the time-frequency diagram of the frequency hopping signal in the Bluetooth frequency band, the square spectrum diagram of the BLE signal, and the square spectrum diagram of the BR / EDR signal. It can be seen that the characteristics of the frequency hopping signal and the two discrete spectral lines on both sides of the image in the BLE square spectrum.

[0119] Experiment 7: Fig.11 It can be seen that the images drawn after autocorrelation operation of FD-LTE and noise signals respectively show the autocorrelation of OFDM modulation compared with irregular noise.

[0120] Experiment 8: Fig.12 It can be seen that the recognition confusion matrix obtained after testing 9 common types of protocol signals and air noise signals in the actual air by the method of this application shows that the recognition accuracy of TD-LTE is over 96%; the recognition accuracy of BLE is over 97%; the recognition accuracy of CDMA2000, FD-LTE, NR, 802.11a / g / n, 802.11b is over 98%; the recognition accuracy of GSM, BR / EDR, Noise is over 99%.

[0121] The signal data set used in this application comes from the collection in real air scenes, and can be used directly as a separate module on the monitoring receiver. The signal is collected through the spectrum monitoring receiver, and the data of multiple transmission protocol models are used as training sets and test sets respectively. By comprehensively considering the complex environment in the air, including the existing noise and interference problems, extracting the characteristics of high robustness of the signal, and combining the deep learning algorithm to fully explore the potential characteristics of the signal, compared with the existing algorithms using simulated data sets, this application not only considers the actual complex scenes, but can also be used on existing monitoring receiver equipment, expanding the functions of the current monitoring receivers, and fully verifying the practicality of the identification method of this application.

[0122] In addition, in response to the problems of few existing wireless transmission protocol recognition algorithms and error accumulation in the algorithms, the recognition method proposed in this application makes full use of the frame structure, modulation method, physical resources and other characteristics involved in each protocol, combined with the autocorrelation of the signal, and through the combination of multimodal features such as time domain, frequency domain, time-frequency domain, and autocorrelation domain, a deep learning algorithm is used to fully utilize the high robustness characteristics of the signal for recognition, thereby improving the accuracy of the recognition method and having stronger robustness and versatility than existing algorithms in complex scenarios.

[0123] The present invention is not limited to the above-mentioned specific embodiments, but extends to any new features or any new combination disclosed in this specification, as well as any new method or process steps or any new combination disclosed.

Claims

1. A signal transmission protocol intelligent identification method for radio monitoring, characterized in that: include: S1. Perform power normalization on the received signal S(n) to obtain a unit power signal S a (n); S2, based on S a The time domain characteristics of (n) determine whether S(n) is a time discrete signal or a time continuous signal; S3, for time-discrete signals, based on S a The high-order spectrum characteristics of S(n) determine whether S(n) is {GSM}, {802.11b}, {CDMA2000} or {NR, TD-LTE, 802.11a / g / n}; if S(n) belongs to {NR, TD-LTE, 802.11a / g / n}, then based on S a The time-frequency characteristics of (n) determine whether S(n) is {NR}, {TD-LTE} or {802.11a / g / n}; S4. For time-continuous signals, based on S a The time-frequency characteristics of S(n) determine whether S(n) is a frequency-hopping signal or a non-frequency-hopping signal; if S(n) is a frequency-hopping signal, then based on S a The high-order spectrum characteristics of S(n) determine whether S(n) is {BR / EDR} or {BLE}; if S(n) is a non-frequency hopping signal, then based on S a The autocorrelation characteristics of S(n) determine whether S(n) is {FD-LTE} or {Noise}.

2. The method for intelligently identifying signal transmission protocols for radio monitoring according to claim 1, characterized in that: The method for normalizing S(n) is: Where L is the number of sampling points of S(n).

3. The signal transmission protocol intelligent identification method for radio monitoring according to claim 1, characterized in that: The step S2 comprises: S2.

1. Obtain S a (n), and sequentially dividing the I / Q sequence into a plurality of I / Q sequence segments; S2.2, sending the I / Q sequence to the stacked dilated convolution module and the first fully connected module for processing to obtain the first feature; S2.3, sending the last I / Q sequence segment to the timing feature convolution module for processing to obtain the second feature; S2.

4. The first feature and the second feature are integrated and sent to the second fully connected module for processing to obtain the classification results of S(n) in the discrete-time signal and the continuous-time signal.

4. The method for intelligently identifying signal transmission protocols for radio monitoring according to claim 3, characterized in that: The stacked dilated convolution module sequentially includes a one-dimensional convolution layer Conv1d, an activation function GELU, a one-dimensional maximum pooling layer MaxPool1d, a one-dimensional convolution layer Conv1d, an activation function GELU, a one-dimensional maximum pooling layer MaxPool1d, a one-dimensional convolution layer Conv1d, an activation function GELU, a one-dimensional maximum pooling layer MaxPool1d, a one-dimensional convolution layer Conv1d, an activation function GELU, a one-dimensional maximum pooling layer MaxPool1d, a one-dimensional convolution layer Conv1d, an activation function GELU, and a one-dimensional maximum pooling layer MaxPool1d; The temporal feature convolution module sequentially includes a one-dimensional convolution layer Conv1d, a one-dimensional average pooling layer AvgPool1d, a KAN network layer KANLayer, a one-dimensional batch normalization layer BatchNorm1d, an activation function ReLU, a one-dimensional convolution layer Conv1d, a one-dimensional average pooling layer AvgPool1d, a KAN network layer KANLayer, an LSTM network layer, an activation function ReLU, a one-dimensional batch normalization layer BatchNorm1d, a KAN network layer KANLayer, and an activation function ReLU; The first fully connected layer and the second fully connected layer each include a linear layer Linear, an activation function ReLU, and a linear layer Linear in sequence.

5. The method for intelligently identifying signal transmission protocols for radio monitoring according to claim 1, characterized in that: In step S3, based on S a The high-order spectral characteristics of (n) determine whether S(n) is {GSM}, {802.11b}, {CDMA2000}, or belongs to {NR, TD-LTE, 802.11a / g / n}, including: S3.1, based on S a The square spectrum characteristics of (n) determine whether S(n) is {GSM}, {802.11b} or belongs to {CDMA2000, NR, TD-LTE, 802.11a / g / n}; S3.2, if S(n) belongs to {CDMA2000, NR, TD-LTE, 802.11a / g / n}, then based on S a The fourth power spectrum characteristics of (n) determine whether S(n) belongs to {CDMA2000} or {NR, TD-LTE, 802.11a / g / n}.

6. The method for intelligently identifying signal transmission protocols for radio monitoring according to claim 1, characterized in that: In step S4, based on S a The high-order spectral characteristics of (n) determine whether S(n) is {BR / EDR} or {BLE}, including: Based on S a The squared spectral characteristics of S(n) determine whether S(n) is {BR / EDR} or {BLE}.

7. The method for intelligently identifying signal transmission protocols for radio monitoring according to claim 5 or 6, characterized in that: In at least one of step S3.1, step S3.2 and step S4, the MobileNet image recognition network is used to extract discrete spectral line information from the square spectrum features or the fourth power spectrum features to classify S(n).

8. The method for intelligently identifying signal transmission protocols for radio monitoring according to claim 7, characterized in that: The MobileNet image recognition network consists of a deep convolution module a, a deep convolution module b and a classification module; The deep convolution module a includes, in sequence, a two-dimensional convolution layer Conv2d, a two-dimensional convolution layer Conv2d, a two-dimensional convolution layer Conv2d, a two-dimensional maximum pooling layer Maxpool2d, an activation function ReLU, a two-dimensional convolution layer Conv2d, a two-dimensional convolution layer Conv2d, a two-dimensional convolution layer Conv2d, a two-dimensional maximum pooling layer Maxpool2d, and an activation function ReLU; The deep convolution module b sequentially includes a two-dimensional convolution layer Conv2d, a two-dimensional batch normalization layer BatchNorm2d, an activation function ReLU, a two-dimensional convolution layer Conv2d, a two-dimensional batch normalization layer BatchNorm2d, an activation function ReLU, a two-dimensional convolution layer Conv2d, a two-dimensional batch normalization layer BatchNorm2d, an activation function ReLU, a two-dimensional convolution layer Conv2d, a two-dimensional batch normalization layer BatchNorm2d, and an activation function ReLU; The classification module includes a two-dimensional average pooling layer Avgpool2d, a flat layer Flatten, a linear layer Linear, an activation function ReLU, a linear layer Linear, an activation function ReLU, and a linear layer Linear in sequence.

9. The method for intelligently identifying signal transmission protocols for radio monitoring according to claim 1, characterized in that: The S a The time-frequency characteristics of (n) are obtained by a (n) is obtained by short-time Fourier transform.

10. A device comprising a processor and a storage medium, wherein the storage medium stores computer instructions, characterized in that: When the processor runs the computer instructions, it can execute the signal transmission protocol intelligent identification method for radio monitoring as described in any one of claims 1-9.

Images