Communication interference signal quantum feature extraction method based on quantum density matrix
By using quantum density matrix transformation and machine/deep learning algorithms to identify communication interference signals, this method solves the problems of insufficient feature extraction accuracy and high complexity in traditional methods, and achieves high-precision interference signal identification under low complexity.
Patent Information
- Application Number
- CN202610103765.0
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-01-26
- Publication Date
- 2026-05-08
AI Technical Summary
Existing methods for identifying communication interference signals suffer from insufficient feature extraction accuracy and high computational complexity, making them difficult to deploy effectively on resource-constrained communication equipment.
A quantum feature extraction method for communication interference signals based on quantum density matrix is adopted. This method converts the time-frequency characteristic matrix into a quantum density matrix and uses machine learning or deep learning algorithms to identify the interference type, including the extraction and identification of TFQDM-1 and TFQDM-2 features.
It achieves high-precision feature extraction of communication interference signals with low complexity, improves the accuracy of interference signal identification, reduces the computational burden, and provides high-resolution feature description.
Smart Images
Figure CN121997023A_ABST
Abstract
Description
Technical Field
[0001] This invention belongs to the field of signal processing technology, specifically relating to a method for extracting quantum features of communication interference signals. Background Technology
[0002] Over the past few decades, the rapid development of wireless communication technology has led to its widespread application in various fields. However, it also faces the severe challenge of an increasingly complex electromagnetic environment. Current communication countermeasures scenarios are characterized by densely distributed radiation sources, a wide frequency range, increased interference power, diversified modulation methods, and the prevalence of low probability of intercept (LCI) signals, making traditional signal detection and processing methods ineffective. Interference sources include both non-human factors such as natural thermal noise and human-caused interference, especially deliberate interference aimed at disrupting communication systems. These interferences take various forms, including deceptive and suppressive, single-tone and multi-tone, narrowband and broadband, posing a comprehensive threat from the air, land, and underwater. This diversity and intelligent evolution of interference makes it difficult for communication systems to decode useful information simply by receiving signals, and even more difficult to guarantee the timeliness of information transmission. Therefore, accurate and rapid intelligent identification of interference signals is the primary prerequisite and core element for implementing effective countermeasure strategies.
[0003] Interference signal identification essentially relies on the quality of feature extraction, the accuracy of which directly determines the reliability of subsequent classification and situational understanding. Traditional methods typically extract handcrafted features from multiple dimensions, including the time, frequency, and spatial domains: the time domain focuses on amplitude / phase transient characteristics and statistics (such as higher-order cumulants and peak-to-average power ratio); the frequency domain analyzes power spectral density distribution, spectral symmetry, and spectral peak positions; and the spatial domain utilizes array signal processing to extract direction of arrival and polarization characteristics. While existing research attempts to automatically mine features using deep learning, it still faces challenges such as high model complexity, reliance on large-scale labeled data, and catastrophic forgetting, making it difficult to deploy on resource-constrained communication devices. Therefore, a new paradigm capable of capturing high-dimensional nonlinear correlation features with low complexity is urgently needed.
[0004] In summary, existing methods still suffer from insufficient feature extraction accuracy and high computational complexity. Faced with the inherent limitations of classical features in terms of dimensionality and information content, researchers are turning to quantum machine learning for breakthroughs. Quantum neural networks, leveraging the Hilbert space of quantum systems, theoretically possess advantages such as exponential storage capacity, simplified network structure, greater stability, and avoidance of catastrophic forgetting, offering entirely new possibilities for high-dimensional signal processing. Summary of the Invention
[0005] The purpose of this invention is to address the problems of insufficient feature extraction accuracy and high computational complexity in existing methods, and to propose a quantum feature extraction method for communication interference signals based on quantum density matrix.
[0006] The technical solution adopted by the present invention to solve the above-mentioned technical problems is as follows:
[0007] According to one aspect of the present invention, a method for extracting quantum features of communication interference signals based on quantum density matrix is provided, the method specifically comprising the following steps:
[0008] Step 1: Acquire the signal to be detected;
[0009] Step 2: Extract the time-frequency characteristic matrix of the signal to be detected;
[0010] Step 3: Convert each column vector of the time-frequency characteristic matrix into a pure state;
[0011] Step 4: Combine the pure states corresponding to each column vector in the time-frequency characteristic matrix to convert the time-frequency characteristic matrix into a quantum density matrix, and use the quantum density matrix as the TFQDM-1 feature of the signal to be detected.
[0012] When identifying interference signals, the TFQDM-1 features of the signal to be detected are used as input to the machine learning algorithm, and the machine learning algorithm outputs the interference type identification result of the signal to be detected.
[0013] Furthermore, the specific process of step three is as follows:
[0014]
[0015] in, Represents the first element in the time-frequency characteristic matrix. column vectors, Representing vectors The 2-norm, Represents the first element in the time-frequency characteristic matrix. The pure state corresponding to each column vector.
[0016] Furthermore, in step four, the pure states corresponding to each column vector in the time-frequency characteristic matrix are combined to convert the time-frequency characteristic matrix into a quantum density matrix. The specific process is as follows:
[0017]
[0018] in, Represents the quantum density matrix. Represents pure state The outer product, This indicates the number of columns in the time-frequency characteristic matrix. Represents pure state The probability of.
[0019] Furthermore, in step four, the probabilities of each pure state satisfy the following: the sum of the probabilities of all pure states is 1, and the probabilities of each pure state are equal.
[0020] The quantum density matrix It is symmetric and positive semi-definite, satisfying the trace of the matrix. .
[0021] According to another aspect of the present invention, a method for extracting quantum features of communication interference signals based on quantum density matrix is provided, the method specifically including the following steps:
[0022] Step 1: Acquire the signal to be detected;
[0023] Step 2: Extract the time-frequency characteristic matrix of the signal to be detected;
[0024] Step 3: Convert each column vector of the time-frequency characteristic matrix into a pure state;
[0025] Step 4: Combine the pure states corresponding to each column vector in the time-frequency characteristic matrix to convert the time-frequency characteristic matrix into a quantum density matrix;
[0026] Step 5: Perform eigenvalue decomposition on the quantum density matrix from Step 4 to obtain the eigenvectors of the quantum density matrix, and then construct a new quantum density matrix based on these eigenvectors. The quantum density matrix TFQDM-2 features of the signal to be detected;
[0027] When identifying interference signals, the TFQDM-2 features of the signal to be detected are used as input to the deep learning algorithm, and the deep learning algorithm outputs the interference type identification result of the signal to be detected.
[0028] Furthermore, the specific process of step three is as follows:
[0029]
[0030] in, Represents the first element in the time-frequency characteristic matrix. column vectors, Representing vectors The 2-norm, Represents the first element in the time-frequency characteristic matrix. The pure state corresponding to each column vector.
[0031] Furthermore, in step four, the pure states corresponding to each column vector in the time-frequency characteristic matrix are combined to convert the time-frequency characteristic matrix into a quantum density matrix. The specific process is as follows:
[0032]
[0033] in, Represents the quantum density matrix. Represents pure state The outer product, This indicates the number of columns in the time-frequency characteristic matrix. Represents pure state The probability of.
[0034] Furthermore, in step four, the probabilities of each pure state satisfy the following: the sum of the probabilities of all pure states is 1, and the probabilities of each pure state are equal.
[0035] quantum density matrix It is symmetric and positive semi-definite, satisfying the trace of the matrix. .
[0036] Furthermore, the quantum density matrix in step four is decomposed into eigenvalues to obtain the eigenvectors of the quantum density matrix, and then a new quantum density matrix is constructed based on these eigenvectors. The quantum density matrix The TFQDM-2 features of the signal to be detected are as follows:
[0037] Step 51: Perform eigenvalue decomposition on the quantum density matrix:
[0038]
[0039] in, Indicates the first 1 eigenvalue, , Indicates the first The eigenvectors corresponding to each eigenvalue. Indicates the number of eigenvectors. Representing the eigenvector The outer product;
[0040] Step 52: Reconstruct the quantum density matrix based on each eigenvector and its corresponding probability to obtain the new quantum density matrix. :
[0041]
[0042] in, Indicates the first The probability corresponding to each feature vector.
[0043] Furthermore, the probabilities corresponding to the feature vectors are obtained based on maximum likelihood estimation.
[0044] The beneficial effects of this invention are:
[0045] This invention first uses a time-frequency characteristic matrix to extract the time-frequency information of the original communication interference signal. Then, it introduces the concept of a quantum density matrix to quantize the time-frequency characteristic matrix to obtain TFQDM-1 features. Next, it uses maximum likelihood estimation on the initially obtained TFQDM-1 features to reconstruct a new quantum density matrix, which is then used as the TFQDM-2 features. The extracted TFQDM-1 and TFQDM-2 features can achieve a computational complexity far lower than that of classical algorithms. At the same time, the extracted features can more comprehensively describe the statistical characteristics of the communication interference signal, improving the accuracy of feature extraction for communication interference signals. The extracted features have advantages that traditional features do not possess, providing a completely new approach to solving problems in classical communication signal processing. Attached Figure Description
[0046] Figure 1 This is a flowchart of a method for extracting quantum features of communication interference signals based on quantum density matrix according to the present invention;
[0047] Figure 2 This is a flowchart of the process for generating the quantum density matrix of communication interference signals;
[0048] Figure 3 This is a framework diagram of a quantum feature extraction method for communication interference signals based on quantum density matrix according to the present invention. Detailed Implementation
[0049] The principle of quantum feature extraction of communication interference signals in this invention is mainly based on the definition of the quantum density matrix. The quantum density matrix describes the superposition of multiple possible quantum states of a complex system and the probability distribution of these quantum states. The probability value of each quantum state reflects the weight of the current sample information. This invention extracts quantum features of communication interference signals based on the quantum density matrix, mapping the classical time-frequency distribution to a high-dimensional quantum state space. By integrating quantum computing, quantum machine learning, and feature extraction techniques for communication interference signals, it provides a high-resolution, low-complexity method for the refined characterization and intelligent identification of communication interference signals in complex electromagnetic environments. The method of this invention will be further described in detail below with reference to the accompanying drawings:
[0050] Specific Implementation Method 1: Combination Figure 1 and Figure 3 This embodiment describes a method for extracting quantum features of communication interference signals based on a quantum density matrix. The method specifically includes the following steps:
[0051] Step 1: Acquire the signal to be detected;
[0052] Step 2: Extract the time-frequency characteristic matrix of the signal to be detected;
[0053] Step 3: To map the classical time-frequency distribution to a high-dimensional quantum state space, the concept of a quantum density matrix is introduced. Each column vector of the time-frequency characteristic matrix is transformed into a pure state, thus linking the processing of communication interference signals with the evolution of quantum states in quantum mechanics. This approach allows each column vector of the time-frequency characteristic matrix to be represented as an orthogonally normalized quantum state; specifically:
[0054] like Figure 2 As shown:
[0055]
[0056] in, Represents the first element in the time-frequency characteristic matrix. column vectors, Representing vectors The 2-norm (also known as the Euclidean norm). Represents the first element in the time-frequency characteristic matrix. The pure state corresponding to each column vector;
[0057] Step 4: Combine the pure states corresponding to each column vector in the time-frequency characteristic matrix to convert the time-frequency characteristic matrix into a quantum density matrix, and use the quantum density matrix as the TFQDM-1 (Time-frequency quantum density matrix) feature of the signal to be detected.
[0058]
[0059] in, Represents the quantum density matrix. The outer product represents the pure state, and each outer product can be viewed as a measurement of the values of some positive operators. This indicates the number of columns in the time-frequency characteristic matrix. Representation based on pure state The subspace formed Represents pure state The probability of;
[0060] It should be noted that:
[0061] 1. The probabilities of each pure state satisfy... In this invention, the probability is set to .
[0062] 2. Quantum density matrix It is symmetric, positive semi-definite, and satisfies ;
[0063] When identifying interference signals, the TFQDM-1 features of the signal to be detected are used as input to a machine learning algorithm, such as, but not limited to, a support vector machine. The machine learning algorithm then outputs the interference type identification result of the signal to be detected.
[0064] Specific Implementation Method Two: The quantum feature extraction method for communication interference signals based on quantum density matrix described in this implementation method specifically includes the following steps:
[0065] Step 1: Acquire the signal to be detected;
[0066] Step 2: Extract the time-frequency characteristic matrix of the signal to be detected;
[0067] Step 3: To map the classical time-frequency distribution to a high-dimensional quantum state space, the concept of a quantum density matrix is introduced. Each column vector of the time-frequency characteristic matrix is transformed into a pure state, thus linking the processing of communication interference signals with the evolution of quantum states in quantum mechanics. This approach allows each column vector of the time-frequency characteristic matrix to be represented as an orthogonally normalized quantum state; specifically:
[0068] like Figure 2 As shown:
[0069]
[0070] in, Represents the first element in the time-frequency characteristic matrix. column vectors, Representing vectors The 2-norm (also known as the Euclidean norm). Represents the first element in the time-frequency characteristic matrix. The pure state corresponding to each column vector;
[0071] Step 4: Combine the pure states corresponding to each column vector in the time-frequency characteristic matrix to convert the time-frequency characteristic matrix into a quantum density matrix;
[0072]
[0073] in, Represents the quantum density matrix. The outer product represents the pure state, and each outer product can be viewed as a measurement of the values of some positive operators. This indicates the number of columns in the time-frequency characteristic matrix. Representation based on pure state The subspace formed Represents pure state The probability of;
[0074] It should be noted that:
[0075] 1. The probabilities of each pure state satisfy... In this invention, the probability is set to .
[0076] 2. Quantum density matrix It is symmetric, positive semi-definite, and satisfies ;
[0077] Step 5: Perform eigenvalue decomposition on the quantum density matrix from Step 4 to obtain the eigenvectors of the quantum density matrix, and then construct a new quantum density matrix based on these eigenvectors. The quantum density matrix TFQDM-2 features of the signal to be detected;
[0078] Step 51: Perform eigenvalue decomposition on the quantum density matrix:
[0079]
[0080] in, Indicates the first 1 eigenvalue, , Indicates the first The eigenvectors corresponding to each eigenvalue. Indicates the number of eigenvectors. Representing the eigenvector The outer product;
[0081] Step 52: Reconstruct the quantum density matrix based on each eigenvector and its corresponding probability to obtain the new quantum density matrix. :
[0082]
[0083] in, Indicates the first The probability corresponding to each feature vector.
[0084] When identifying interference signals, the TFQDM-2 features of the signal to be detected are used as input to a deep learning algorithm, including but not limited to a deep belief network. The deep learning algorithm then outputs the interference type identification result of the signal to be detected.
[0085] Experimental section:
[0086] We selected three types of communication interference signals (single-tone interference, linear sweep frequency interference, and periodic impulse noise interference) to describe the feature extraction process. First, we simulated these three interference signals using simulation software. Then, we read the original interference signal sample data and converted it into a time-frequency characteristic matrix. Next, we used the method in step four to convert the time-frequency characteristic matrix into TFQDM-1 features. Finally, we performed feature decomposition on the TFQDM-1 features and then performed maximum likelihood estimation to obtain the probability of each feature vector obtained from the feature decomposition. Specifically:
[0087] In quantum mechanics, the density matrix describes the quantum state of a system, while the maximum likelihood state is the most likely quantum state to describe the system given the density matrix.
[0088] The most probable probability distribution corresponding to each eigenvector is obtained through maximum likelihood estimation (i.e., the probability corresponding to each eigenvector is obtained separately), and the obtained probability distribution is normalized to obtain the new probability corresponding to each eigenvector. The new probabilities corresponding to the feature vectors are denoted as . ;
[0089] The new quantum density matrix is obtained by reconstructing the matrix based on each eigenvector and its corresponding new probability. :
[0090]
[0091] The new quantum density matrix TFQDM-2 features are used as sample data.
[0092] To verify the effectiveness of TFQDM features, we applied them to a communication interference signal identification task for evaluation. The obtained TFQDM-1 and TFQDM-2 features were used as inputs to a support vector machine (SVM), and the interference identification accuracy based on TFQDM-1 features and TFQDM-2 features was obtained from the SVM output, as shown in Table 1.
[0093] Table 1. Accuracy of Identification of Communication Interference Signals of Various Characteristics
[0094]
[0095] It is evident that the highest accuracy can be achieved when using machine learning algorithms based on TFQDM-1 features. Therefore, machine learning algorithms should be selected when using TFQDM-1 features for identification.
[0096] For deep learning algorithms, we chose the commonly used Deep Belief Network (DBN) to perform the communication interference signal identification task. During the training process, this invention trained corresponding deep belief networks based on the time-frequency characteristic matrix, TFQDM-1 features, and TFQDM-2 features, respectively. Except for a few parameters that need to be adjusted due to the input feature dimension, the other model parameters of the three deep belief networks remained unchanged. We input the time-frequency characteristic matrix, TFQDM-1, and TFQDM-2 features into the corresponding trained deep belief networks for identification. The identification accuracy of each feature is shown in Table 2. From Table 2, we can see that in the identification tasks of these three interference signals, TFQDM features are better than the classic time-frequency features. Compared with TFQDM-1 features, TFQDM-2 features can reduce the feature dimension while maximizing the preservation of sample feature information through feature decomposition and maximum likelihood estimation, thereby reducing the computational burden of the model and improving the identification accuracy of the DBN model for interference signals.
[0097] Table 2. Accuracy of Identification of Communication Interference Signals of Various Characteristics
[0098]
[0099] Therefore, when using deep learning algorithms, the highest accuracy can be achieved based on TFQDM-2 features. When using TFQDM-2 features for identification, a deep learning algorithm should be chosen. When training the DBN model using interference signal sample data, the model parameters are continuously adjusted to obtain optimal parameters. In actual deployment, the input TFQDM-2 features need to be identified based on the obtained optimal model parameters.
[0100] In summary, at the quantum feature extraction level, the quantum density matrix is used as a complete description of the signal's quantum state, mapping the classical time-frequency distribution to a high-dimensional quantum state space. Quantum parallelism and entanglement effects are utilized to capture cross-domain correlated TFQDM features in a single operation. These TFQDM features are then applied to the identification of communication interference signals. As quantum features extracted from the classical time-frequency characteristic matrix, TFQDM features exhibit excellent discriminative power. Besides effectively identifying communication interference signals, this approach integrates quantum computing, quantum machine learning, and feature extraction techniques for communication interference signals, resulting in high-resolution, low-complexity TFQDM features compared to traditional features.
[0101] The above examples of the present invention are merely illustrative of the computational model and process of the present invention, and are not intended to limit the implementation of the present invention. Those skilled in the art will recognize that other variations or modifications can be made based on the above description. It is impossible to exhaustively list all possible implementations here. Any obvious variations or modifications derived from the technical solutions of the present invention are still within the scope of protection of the present invention.
Claims
1. A method for extracting quantum features of communication interference signals based on quantum density matrix, characterized in that, The method specifically includes the following steps: Step 1: Acquire the signal to be detected; Step 2: Extract the time-frequency characteristic matrix of the signal to be detected; Step 3: Convert each column vector of the time-frequency characteristic matrix into a pure state; Step 4: Combine the pure states corresponding to each column vector in the time-frequency characteristic matrix to convert the time-frequency characteristic matrix into a quantum density matrix, and use the quantum density matrix as the TFQDM-1 feature of the signal to be detected. When identifying interference signals, the TFQDM-1 features of the signal to be detected are used as input to the machine learning algorithm, and the machine learning algorithm outputs the interference type identification result of the signal to be detected.
2. The method for extracting quantum features of communication interference signals based on quantum density matrix according to claim 1, characterized in that, The specific process of step three is as follows: in, Represents the first element in the time-frequency characteristic matrix. column vectors, Representing vectors The 2-norm, Represents the first element in the time-frequency characteristic matrix. The pure state corresponding to each column vector.
3. The method for extracting quantum features of communication interference signals based on quantum density matrix according to claim 2, characterized in that, In step four, the pure states corresponding to each column vector in the time-frequency characteristic matrix are combined to convert the time-frequency characteristic matrix into a quantum density matrix. The specific process is as follows: in, Represents the quantum density matrix. Represents pure state The outer product, This indicates the number of columns in the time-frequency characteristic matrix. Represents pure state The probability of.
4. The method for extracting quantum features of communication interference signals based on quantum density matrix according to claim 3, characterized in that, In step four, the probability of each pure state satisfies the following: the sum of the probabilities of all pure states is 1, and the probabilities of each pure state are equal. The quantum density matrix It is symmetric and positive semi-definite, satisfying the trace of the matrix. .
5. A method for extracting quantum features of communication interference signals based on quantum density matrix, characterized in that, The method specifically includes the following steps: Step 1: Acquire the signal to be detected; Step 2: Extract the time-frequency characteristic matrix of the signal to be detected; Step 3: Convert each column vector of the time-frequency characteristic matrix into a pure state; Step 4: Combine the pure states corresponding to each column vector in the time-frequency characteristic matrix to convert the time-frequency characteristic matrix into a quantum density matrix; Step 5: Perform eigenvalue decomposition on the quantum density matrix from Step 4 to obtain the eigenvectors of the quantum density matrix, and then construct a new quantum density matrix based on these eigenvectors. The quantum density matrix TFQDM-2 features of the signal to be detected; When identifying interference signals, the TFQDM-2 features of the signal to be detected are used as input to the deep learning algorithm, and the deep learning algorithm outputs the interference type identification result of the signal to be detected.
6. The method for extracting quantum features of communication interference signals based on quantum density matrix according to claim 5, characterized in that, The specific process of step three is as follows: in, Represents the first element in the time-frequency characteristic matrix. column vectors, Representing vectors The 2-norm, Represents the first element in the time-frequency characteristic matrix. The pure state corresponding to each column vector.
7. The method for extracting quantum features of communication interference signals based on quantum density matrix according to claim 6, characterized in that, In step four, the pure states corresponding to each column vector in the time-frequency characteristic matrix are combined to convert the time-frequency characteristic matrix into a quantum density matrix. The specific process is as follows: in, Represents the quantum density matrix. Represents pure state The outer product, This indicates the number of columns in the time-frequency characteristic matrix. Represents pure state The probability of.
8. The method for extracting quantum features of communication interference signals based on quantum density matrix according to claim 7, characterized in that, In step four, the probability of each pure state satisfies the following: the sum of the probabilities of all pure states is 1, and the probabilities of each pure state are equal. The quantum density matrix It is symmetric and positive semi-definite, satisfying the trace of the matrix. .
9. The method for extracting quantum features of communication interference signals based on quantum density matrix according to claim 8, characterized in that, The quantum density matrix in step four is decomposed into eigenvalues to obtain the eigenvectors of the quantum density matrix. A new quantum density matrix is then constructed based on these eigenvectors. The quantum density matrix The TFQDM-2 features of the signal to be detected are as follows: Step 51: Perform eigenvalue decomposition on the quantum density matrix: in, Indicates the first 1 eigenvalue, , Indicates the first The eigenvectors corresponding to each eigenvalue. Indicates the number of eigenvectors. Representing the eigenvector The outer product; Step 52: Reconstruct the quantum density matrix based on each eigenvector and its corresponding probability to obtain the new quantum density matrix. : in, Indicates the first The probability corresponding to each feature vector.
10. The method for extracting quantum features of communication interference signals based on quantum density matrix according to claim 9, characterized in that, The probabilities corresponding to the feature vectors are obtained based on maximum likelihood estimation.