Millimeter wave radar arrhythmia detection method based on particle swarm optimization variational mode decomposition and deep learning

Abstract

The invention belongs to the field of non-contact vital sign detection, and particularly relates to a millimeter wave radar arrhythmia detection method based on particle swarm optimization variational mode decomposition and deep learning, and the method specifically comprises the steps: S1, obtaining a human chest micro-motion signal through a millimeter wave radar, performing static clutter filtering, phase extraction, unwrapping and detrending processing on the radar echo signal to obtain a chest displacement signal containing heartbeat and breathing information; s2, aiming at the thoracic cavity displacement signal, constructing a variational mode decomposition model, and carrying out adaptive optimization on a mode number and a penalty factor through a particle swarm optimization algorithm to obtain an optimal decomposition parameter and complete signal decomposition; s3, according to the center frequency and the energy distribution characteristics of each modal component, screening the modal components in the heartbeat frequency range and reconstructing the modal components to obtain heartbeat characteristic signals representing heart mechanical activities; s4, carrying out time sequence segmentation on the heartbeat characteristic signals, extracting local heart beat morphological characteristics by utilizing a convolutional neural network, and carrying out modeling on a long-time rhythm dependency relationship of the heartbeat signals in combination with a time sequence modeling network based on an attention mechanism to obtain heart rhythm depth characteristic representation; and S5, inputting the heart rhythm depth features into a heart rhythm discrimination model, analyzing the heart rhythm state of the detected person, and outputting an arrhythmia detection result. According to the method, adaptive selection of variational mode decomposition parameters is realized by introducing a particle swarm optimization mechanism, heartbeat signals and respiration and motion interference components are effectively separated, modeling is carried out on rhythm characteristics in combination with a deep learning model, and non-contact detection of arrhythmia is realized. The method does not need to wear an electrode or contact a human body, has the advantages of strong anti-interference capability, good adaptability and high detection precision, and has a good application prospect in the fields of heart rhythm health monitoring, disease screening and the like.

Description

Technical Field

[0001] This invention belongs to the field of non-contact vital sign detection technology, specifically a millimeter-wave radar arrhythmia detection method based on particle swarm optimization variational mode decomposition and deep learning. Background Technology

[0002] Cardiac arrhythmias are a common group of abnormal heart rhythms, including atrial fibrillation, irregular heartbeats, tachycardia, and bradycardia. In severe cases, they can lead to complications such as stroke and heart failure. Current methods for detecting cardiac arrhythmias mainly rely on contact-based detection methods such as electrocardiograms (ECG), which require electrodes to be in direct contact with the skin. This presents inconvenience in long-term continuous monitoring, sleep monitoring, and special scenarios such as skin injury or burns.

[0003] In recent years, millimeter-wave radar has been widely used in the field of human vital sign detection due to its advantages such as non-contact operation, good privacy, and ability to penetrate clothing. Millimeter-wave radar can acquire radar echo signals containing information about breathing and heartbeat by sensing subtle movements of the human chest cavity. However, in practical applications, unavoidable minor body movements, respiratory interference, and environmental noise can severely affect the accuracy of heartbeat signal extraction, thereby reducing the reliability of heart rhythm analysis.

[0004] Existing heart rhythm detection methods based on millimeter-wave radar mostly employ fixed-parameter filtering or traditional signal decomposition methods to extract heartbeat signals. However, these methods often struggle to balance signal separation effectiveness with parameter adaptability when faced with signal non-stationarity and individual differences. Furthermore, some methods rely solely on simple features such as heart rate or heart rate variability for discrimination, making it difficult to effectively characterize complex rhythmic changes.

[0005] Therefore, a non-contact arrhythmia detection method that can adaptively separate heartbeat signals and perform in-depth modeling of heart rhythm is needed to improve the accuracy and stability of detection. Summary of the Invention

[0006] (a) Technical problems to be solved

[0007] To address the shortcomings of existing technologies, this invention proposes a millimeter-wave radar arrhythmia detection method based on particle swarm optimization variational mode decomposition and deep learning, which solves the problems mentioned in the background technology.

[0008] (II) Technical Solution

[0009] To achieve the above objectives, the present invention specifically adopts the following technical solution:

[0010] A millimeter-wave radar arrhythmia detection method based on particle swarm optimization variational mode decomposition and deep learning includes the following steps;

[0011] S1. Obtain the original echo signal matrix acquired by the millimeter-wave radar from the millimeter-wave radar echo signal dataset. Since the original echo signal contains reflections from fixed environmental targets and system noise, static clutter filtering is performed on the original echo signal matrix to reduce these interferences. Subsequently, phase extraction is performed on the radar echo signal, and the extracted phase signal is unwrapped to eliminate the influence of phase jumps on the measurement of minute displacements. Further, detrending processing is used to remove ultra-low frequency components caused by slow human body movement or system drift, thereby obtaining the chest displacement signal that mainly reflects the micro-motion information of the human chest cavity. This provides a stable input for subsequent mode decomposition.

[0012] S2, using variational mode decomposition to solve the following variational optimization problem, including the following steps:

[0013] S201, regarding the thoracic displacement signal obtained in step S1 A variational mode decomposition model is constructed. The core idea of ​​variational mode decomposition is to ensure that the linear superposition of each mode component can completely reconstruct the original signal, while making each mode component exhibit the minimum bandwidth near its corresponding center frequency, thereby achieving adaptive frequency band separation for non-stationary signals.

[0014] Its optimization model is expressed as:

[0015]

[0016]

[0017] in, This represents the k-th modal component. This represents the corresponding center frequency. By introducing an analytic signal form and performing frequency shifting on the modes, the above objective function can effectively characterize the bandwidth characteristics of the modes in the frequency domain.

[0018] S202, To solve the above constrained optimization problem, Lagrange multipliers are introduced. and punishment factors Transform it into an unconstrained augmented Lagrange form:

[0019]

[0020] By transforming the above expression to the frequency domain, the modal bandwidth term can be converted into a frequency domain equivalent to the frequency domain. This provides an equivalent form, facilitating efficient solution using the Fast Fourier Transform.

[0021] S203 uses the alternating direction multiplier method to solve the augmented Lagrangian function. By fixing some variables and updating the remaining variables one by one, the original problem is broken down into several easily solvable subproblems.

[0022] In the nth iteration, the update formula for the kth modal component in the frequency domain is:

[0023]

[0024] This updated formula achieves adaptive suppression and retention of different frequency components by weighting the frequency domain energy.

[0025] Correspondingly, the center frequency is updated using an energy-weighted average method:

[0026]

[0027] The above update method enables the center frequency to gradually converge towards the frequency band where the main energy of the mode is concentrated.

[0028] The Lagrange multipliers are updated as follows:

[0029]

[0030] When the preset convergence condition is met or the maximum number of iterations is reached, the iteration stops and the final mode decomposition result is obtained.

[0031] S3 uses a particle swarm optimization algorithm to iteratively update parameters, so that while meeting the signal reconstruction accuracy, the intermodal frequency separation and heartbeat energy concentration are optimized, thereby improving the decomposition stability under different individuals and different measurement conditions.

[0032] Given that the number of modes K and the penalty factor α have a significant impact on the decomposition effect, the parameter vector is defined as follows:

[0033]

[0034] The modal components obtained from variational mode decomposition are screened based on their center frequency and energy distribution characteristics. First, a frequency threshold is set according to the physiological heart rate range:

[0035]

[0036] Furthermore, the modal energy ratio is introduced as a constraint:

[0037]

[0038] in: and ;

[0039] By using the above dual criteria, respiratory components and noise modes can be effectively excluded, thus improving the reliability of heartbeat mode determination.

[0040] S4, according to the method described in claim S3, characterized in that: the heartbeat characteristic signal is reconstructed by the following formula:

[0041]

[0042] in: This represents the set of modality indices that satisfy the heartbeat modality screening criteria;

[0043] S5, using a temporal modeling network based on convolutional neural networks and attention mechanisms to determine heart rhythm status, includes the following steps:

[0044] S501, transmits heartbeat characteristic signals Normalization was performed, and the samples were segmented according to the time window length T to construct heartbeat time series samples;

[0045] S502, the heartbeat timing sample is input into a convolutional neural network to extract local heartbeat morphological features of the heartbeat signal, the feature mapping of which is expressed as:

[0046]

[0047] S503, The local features are input into an attention-based temporal modeling network to calculate attention weights and obtain heart rate depth features:

[0048]

[0049] S504, Calculate the discrimination probability corresponding to each heart rhythm category based on the heart rhythm depth features:

[0050]

[0051] S505 outputs the heart rhythm status of the tested person according to the principle of maximum discrimination probability and determines whether there is arrhythmia.

[0052] (III) Beneficial Effects

[0053] Compared with existing technologies, this invention provides a millimeter-wave radar arrhythmia detection method based on particle swarm optimization variational mode decomposition and deep learning, which has the following beneficial effects:

[0054] This method achieves adaptive separation of the heartbeat component in millimeter-wave radar thoracic displacement signals by constructing a variational mode decomposition model based on particle swarm optimization. Through adaptive optimization of the variational mode decomposition parameters, combined with a rigorous variational optimization and iterative solution mechanism, different physiological components are effectively distinguished in the frequency domain. This reduces the impact of respiratory interference and body motion noise on heartbeat signal extraction, improving the stability and reliability of heartbeat feature signal extraction.

[0055] This invention combines the reconstructed heartbeat feature signals with a deep learning model to jointly model and discriminate heart rhythms. By comprehensively modeling the local cardiac morphological features and long-term rhythm dependencies of the heartbeat signal, it achieves non-contact detection of arrhythmias, improving the accuracy and applicability of arrhythmia discrimination. Attached Figure Description

[0056] Figure 1 This is a flowchart of a millimeter-wave radar arrhythmia detection method based on particle swarm optimization variational mode decomposition and deep learning according to the present invention.

[0057] Figure 2 This is the time-domain plot of the signal corresponding to the intrinsic mode functions after decomposition using this method;

[0058] Figure 3 This is the frequency domain diagram of the signal corresponding to the intrinsic mode functions after decomposition using this method;

[0059] Figure 4 Waveforms of respiratory and heartbeat signals; Detailed Implementation

[0060] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0061] Example

[0062] like Figure 1 As shown, an embodiment of the present invention proposes a millimeter-wave radar arrhythmia detection method based on particle swarm optimization variational mode decomposition and deep learning, comprising the following steps:

[0063] S1. Obtain the original echo signal matrix acquired by the millimeter-wave radar from the millimeter-wave radar echo signal dataset. Since the original echo signal contains reflections from fixed environmental targets and system noise, static clutter filtering is performed on the original echo signal matrix to reduce these interferences. Subsequently, phase extraction is performed on the radar echo signal, and the extracted phase signal is unwrapped to eliminate the influence of phase jumps on the measurement of minute displacements. Further, detrending processing is used to remove ultra-low frequency components caused by slow human body movement or system drift, thereby obtaining the chest displacement signal that mainly reflects the micro-motion information of the human chest cavity. This provides a stable input for subsequent mode decomposition.

[0064] S2, using variational mode decomposition to solve the following variational optimization problem, including the following steps:

[0065] S201, regarding the thoracic displacement signal obtained in step S1 A variational mode decomposition model is constructed. The core idea of ​​variational mode decomposition is to ensure that the linear superposition of each mode component can completely reconstruct the original signal, while making each mode component exhibit the minimum bandwidth near its corresponding center frequency, thereby achieving adaptive frequency band separation for non-stationary signals.

[0066] Its optimization model is expressed as:

[0067]

[0068]

[0069] in, This represents the k-th modal component. This represents the corresponding center frequency. By introducing an analytic signal form and performing frequency shifting on the modes, the above objective function can effectively characterize the bandwidth characteristics of the modes in the frequency domain.

[0070] S202, To solve the above constrained optimization problem, Lagrange multipliers are introduced. and punishment factors Transform it into an unconstrained augmented Lagrange form:

[0071]

[0072] By transforming the above expression to the frequency domain, the modal bandwidth term can be converted into a frequency domain equivalent to the frequency domain. This provides an equivalent form, facilitating efficient solution using the Fast Fourier Transform.

[0073] S203 uses the alternating direction multiplier method to solve the augmented Lagrangian function. By fixing some variables and updating the remaining variables one by one, the original problem is broken down into several easily solvable subproblems.

[0074] In the nth iteration, the update formula for the kth modal component in the frequency domain is:

[0075]

[0076] This updated formula achieves adaptive suppression and retention of different frequency components by weighting the frequency domain energy.

[0077] Correspondingly, the center frequency is updated using an energy-weighted average method:

[0078]

[0079] The above update method enables the center frequency to gradually converge towards the frequency band where the main energy of the mode is concentrated.

[0080] The Lagrange multipliers are updated as follows:

[0081]

[0082] When the preset convergence condition is met or the maximum number of iterations is reached, the iteration stops and the final mode decomposition result is obtained.

[0083] S3 uses a particle swarm optimization algorithm to iteratively update parameters, so that while meeting the signal reconstruction accuracy, the intermodal frequency separation and heartbeat energy concentration are optimized, thereby improving the decomposition stability under different individuals and different measurement conditions.

[0084] Given that the number of modes K and the penalty factor α have a significant impact on the decomposition effect, the parameter vector is defined as follows:

[0085]

[0086] The modal components obtained from variational mode decomposition are screened based on their center frequency and energy distribution characteristics. First, a frequency threshold is set according to the physiological heart rate range:

[0087]

[0088] Furthermore, the modal energy ratio is introduced as a constraint:

[0089]

[0090] in: and ;

[0091] By using the above dual criteria, respiratory components and noise modes can be effectively excluded, thus improving the reliability of heartbeat mode determination.

[0092] S4, according to the method described in claim S3, characterized in that: the heartbeat characteristic signal is reconstructed by the following formula:

[0093]

[0094] in: This represents the set of modality indices that satisfy the heartbeat modality screening criteria;

[0095] S5, using a temporal modeling network based on convolutional neural networks and attention mechanisms to determine heart rhythm status, includes the following steps:

[0096] S501, transmits heartbeat characteristic signals Normalization was performed, and the samples were segmented according to the time window length T to construct heartbeat time series samples;

[0097] S502, the heartbeat timing sample is input into a convolutional neural network to extract local heartbeat morphological features of the heartbeat signal, the feature mapping of which is expressed as:

[0098]

[0099] S503, The local features are input into an attention-based temporal modeling network to calculate attention weights and obtain heart rate depth features:

[0100]

[0101] S504, Calculate the discrimination probability corresponding to each heart rhythm category based on the heart rhythm depth features:

[0102]

[0103] S505 outputs the heart rhythm status of the tested person according to the principle of maximum discrimination probability and determines whether there is arrhythmia.

[0104] This invention targets one-dimensional physiological signals of chest cavity micro-motion acquired by millimeter-wave radar. It employs a variational mode decomposition method based on particle swarm optimization to adaptively decompose the signal. By optimizing the decomposition parameters, the heartbeat component is effectively separated in the frequency domain. Combining the physiological frequency range of the heartbeat and the modal energy distribution characteristics, the decomposed modal components are screened and the heartbeat characteristic signal is reconstructed. This not only effectively suppresses respiratory interference and motion noise but also preserves the temporal morphology and rhythmic characteristics of the heartbeat signal. Based on this, the reconstructed heartbeat characteristic signal is input into a deep learning model to model and discriminate the heart rhythm, concentrating the heartbeat energy within a reasonable physiological frequency range. This improves the quality of millimeter-wave radar vital sign signals and provides technical support for subsequent applications in medical monitoring, health detection, and emergency rescue.

[0105] Finally, it should be noted that the above descriptions are merely preferred embodiments of the present invention and are not intended to limit the invention. For those skilled in the art, modifications can be made to the technical solutions described in the above embodiments, or equivalent substitutions can be made to some of the technical features, without departing from the spirit and principles of the present invention. Any modifications, equivalent substitutions, or improvements made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A millimeter-wave radar arrhythmia detection method based on particle swarm optimization variational mode decomposition and deep learning, characterized in that, The method includes the following steps: S1: Obtain the original echo signal matrix from the millimeter-wave radar echo signal dataset, and perform static clutter filtering, phase extraction, unwrapping and detrending processing on the original echo signal matrix to obtain a thoracic displacement signal containing micro-motion information of the human thoracic cavity. S2: Construct a variational mode decomposition model for the chest cavity displacement signal, and introduce a particle swarm optimization algorithm to adaptively optimize the number of modes and the penalty factor in the variational mode decomposition model to obtain the optimal decomposition parameters. Then, use the optimal decomposition parameters to perform mode decomposition on the chest cavity displacement signal to obtain several mode components with different center frequencies. S3: Based on the center frequency and energy distribution characteristics of the modal components, filter the modal components within the preset heartbeat frequency range, and reconstruct the heartbeat modal components to obtain the heartbeat characteristic signal that characterizes the mechanical activity of the human heart. S4: The heartbeat feature signal is segmented according to a time window, and a deep learning model is used to jointly model the local heartbeat morphology features and their long-term rhythm dependence in the heartbeat feature signal to extract the heart rhythm depth features that represent the heart rhythm state. S5: Input the heart rhythm depth features into the heart rhythm discrimination model to analyze and classify the heart rhythm status of the tested person and output the heart rhythm detection results.

2. The millimeter-wave radar arrhythmia detection method based on particle swarm optimization variational mode decomposition and deep learning according to claim 1, characterized in that: The variational mode decomposition is achieved by solving the following variational optimization problem: in: This indicates the thoracic cavity displacement signal. Indicates the first One modal component, This indicates the corresponding center frequency.

3. The millimeter-wave radar arrhythmia detection method based on particle swarm optimization variational mode decomposition and deep learning according to claim 1, characterized in that: Particle swarm optimization algorithm uses variational mode decomposition of parameter vectors To optimize the variables, the optimal parameter combination is obtained through particle swarm optimization. The heartbeat modal components The following screening criteria must be met: in: and .

4. The method according to claim 3, characterized in that: The heartbeat characteristic signal is reconstructed using the following formula: in: This represents the set of modality indices that satisfy the heartbeat modality screening criteria.

5. The millimeter-wave radar arrhythmia detection method based on particle swarm optimization variational mode decomposition and deep learning according to claim 1, characterized in that: The heart rhythm state is determined using a temporal modeling network based on convolutional neural networks and attention mechanisms, including the following steps: S501, transmits heartbeat characteristic signals Normalization was performed, and the samples were segmented according to the time window length T to construct heartbeat time series samples; S502, the heartbeat timing sample is input into a convolutional neural network to extract local heartbeat morphological features of the heartbeat signal, the feature mapping of which is expressed as: S503, The local features are input into an attention-based temporal modeling network to calculate attention weights and obtain heart rate depth features: S504, Calculate the discrimination probability corresponding to each heart rhythm category based on the heart rhythm depth features: S505 outputs the heart rhythm status of the tested person according to the principle of maximum discrimination probability and determines whether there is arrhythmia.

Images