BCG heart rate detection device based on terahertz radar

Abstract

The invention relates to the technical field of heart rate detection, in particular to a BCG heart rate detection device based on a terahertz radar, and the device comprises a multi-channel terahertz radar module which is used for transmitting electromagnetic waves to a plurality of to-be-detected targets at the same time; meanwhile, receiving reflected electromagnetic waves, and performing signal preprocessing and signal decomposition on the electromagnetic waves to obtain two paths of I / Q orthogonal echo signals; the signal processing module is used for carrying out signal preprocessing on the I / Q two-path orthogonal echo signals to obtain processed signals; the heart rate calculation module is used for inputting the processed signal into a heart rate feature algorithm model to obtain a final heart rate result; the heart rate feature algorithm model comprises a heart rate feature extraction sub-model and a heart rate acquisition sub-model; the heart rate acquisition sub-model is a regression model based on a generative adversarial network. The method provided by the invention solves the technical problem of insufficient monitoring precision and stability in the prior art.

Description

Technical Field

[0001] The present invention relates to the technical field of heart rate detection, and in particular, to a BCG heart rate detection device based on terahertz radar. Background Art

[0002] With the development of terahertz technology, terahertz through-wall radar, as a new type of detection technology, has been applied in many fields. Terahertz has the characteristics of strong penetrability and high resolution, making it have a wide range of application scenarios in fields such as security inspection and medical diagnosis. At present, terahertz through-wall radar technology is mainly applied to the imaging of objects with millimeter-level accuracy, and the detection and analysis of dynamic targets are still in the initial stage.

[0003] Currently, the heart rate monitoring of terahertz radar mainly adopts a single-point monitoring method, which is difficult to achieve real-time monitoring and analysis of multiple targets. At the same time, in the existing technology, when monitoring at a long distance, the signal attenuation is serious and it is easily interfered by environmental factors, resulting in insufficient monitoring accuracy and stability.

[0004] Therefore, there is an urgent need for a BCG heart rate detection device based on terahertz radar. Summary of the Invention

[0005] (I) Technical Problems to be Solved

[0006] In view of the above-mentioned shortcomings and deficiencies of the prior art, the present invention provides a BCG heart rate detection device based on terahertz radar, which solves the technical problems of insufficient monitoring accuracy and stability in the prior art.

[0007] (II) Technical Solutions

[0008] In order to achieve the above object, the main technical solutions adopted by the present invention include:

[0009] An embodiment of the present invention provides a BCG heart rate detection device based on terahertz radar, including:

[0010] A multi-channel terahertz radar module, configured to simultaneously transmit electromagnetic waves to multiple targets to be measured through a MIMO antenna array; simultaneously receive the reflected electromagnetic waves, and perform signal preprocessing and signal decomposition on the electromagnetic waves to obtain I / Q two-channel quadrature echo signals;

[0011] A signal processing module, configured to perform signal preprocessing on the I / Q two-channel quadrature echo signals to obtain processed signals;

[0012] A heart rate calculation module, configured to input the processed signals into a heart rate feature algorithm model to obtain a final heart rate result; the heart rate feature algorithm model includes a heart rate feature extraction sub-model and a heart rate acquisition sub-model; the heart rate acquisition sub-model is a regression model based on a generative adversarial network.

[0013] Optionally, the multi-channel terahertz radar module includes:

[0014] A transmitting module and a receiving module;

[0015] The transmitting module is used to mix the baseband signal with the local oscillator signal to the terahertz frequency band and transmit it through the MIMO antenna array;

[0016] The receiving module includes a beamforming unit, a down-conversion processing unit and a decomposition unit. The beamforming unit is used to separate multiple targets from the received reflected signal, and the decomposition unit is used to decompose the down-converted baseband signal into I / Q two-channel quadrature echo signals.

[0017] Optionally, the MIMO antenna array is an 8×8 planar array, the adjacent antenna spacing is 0.8 - 1.2 times the operating wavelength, and the operating frequency band is 0.3 - 0.5 THz.

[0018] Optionally, the signal processing module includes:

[0019] A denoising unit that filters the I / Q two-channel quadrature echo signals in the 0.5 - 4 Hz frequency band using an adaptive band-pass filter to obtain the first I / Q two-channel quadrature echo signals;

[0020] A wavelet transform unit that performs 3-layer decomposition using the Daubechies 4 wavelet basis, retains the coefficients in the 0.5 - 4 Hz frequency band, and obtains the second I / Q two-channel quadrature echo signals;

[0021] A correction unit that performs 5-layer modal separation on the respiratory harmonic interference through variational mode decomposition, removes the first 3 layers of low-frequency respiratory components, and obtains the I / Q two-channel quadrature echo signals without respiratory harmonic interference as the processed signals.

[0022] Optionally, the heart rate feature algorithm model includes:

[0023] A heart rate feature extraction sub-model that extracts the time-frequency features of the signal through multi-scale wavelet decomposition and dilated convolution;

[0024] A heart rate acquisition sub-model that inputs the time-frequency features into the pre-constructed heart rate acquisition sub-model to obtain the final heart rate result.

[0025] Optionally, the heart rate feature extraction sub-model includes:

[0026] An input layer that inputs the processed signal;

[0027] A wavelet transform decomposition layer that performs 4-layer decomposition using the Morlet wavelet basis, covering the 0.5 - 4 Hz frequency band, to obtain signals of different scales;

[0028] A convolutional neural network layer is used to perform feature extraction on the signal at each scale to capture the local and global features of the signal; the convolutional neural network layer uses a one-dimensional convolution layer with a kernel size of 5 and a dilation factor of 3.

[0029] The feature fusion layer is used to obtain the attention weights of multi-scale features through SENet and obtain the weighted feature vector;

[0030] The output layer outputs weighted heart rate features including heart rate variability and rhythmic patterns.

[0031] Optionally, the heart rate acquisition sub-model includes a generator and a discriminator;

[0032] The generator is a 3-layer LSTM network with a hidden layer dimension of 128;

[0033] The discriminator is a 4-layer one-dimensional convolutional network.

[0034] Optionally, the heart rate acquisition sub-model is a sub-model trained using a training data set;

[0035] The training data set includes real heart rate signals of different historical environmental conditions and different target individuals, as well as corresponding heart rate variability and rhythm patterns.

[0036] Optionally, when training the heart rate acquisition sub-model, the generator and the discriminator are trained alternately, the generator loss function is a weighted sum of a mean square error and an adversarial loss, and the discriminator loss function is a cross entropy loss;

[0037] The training is terminated when the generator loss function decreases by less than 1% in 10 consecutive iterations, and the decrease rate is calculated by the sliding window averaging method.

[0038] Optionally, in the correction unit, the number of modes of variational modal decomposition K=6, the penalty factor α=2000, and the iteration stop condition is that the modal energy change rate ≤1%.

[0039] (III) Beneficial effects

[0040] The beneficial effects of the present invention are as follows: a BCG heart rate detection device based on terahertz radar of the present invention adopts a multi-channel terahertz radar module, simultaneously transmits electromagnetic waves to multiple targets to be measured through a MIMO antenna array, and uses a heart rate acquisition sub-model based on a generative adversarial network, thereby realizing real-time heart rate detection of multiple targets and improving the accuracy of heart rate detection, thereby improving the reliability and practicality of the detection results. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] Figure 1Schematic structural diagram of a BCG heart rate detection device based on terahertz radar according to an embodiment of the present invention. Detailed implementation manners

[0042] In order to better explain the present invention for easy understanding, the present invention will be described in detail below with reference to the accompanying drawings through specific implementation manners.

[0043] In order to better understand the above technical solution, the exemplary embodiments of the present invention will be described in more detail below with reference to the accompanying drawings. Although the exemplary embodiments of the present invention are shown in the drawings, it should be understood that the present invention can be implemented in various forms and should not be limited by the embodiments set forth herein. On the contrary, these embodiments are provided to enable a more clear and thorough understanding of the present invention and to fully convey the scope of the present invention to those skilled in the art.

[0044] Embodiment 1

[0045] See Figure 1 , a BCG heart rate detection device based on terahertz radar in this embodiment includes:

[0046] A multi-channel terahertz radar module, configured to simultaneously transmit electromagnetic waves to multiple targets to be measured through a MIMO antenna array; simultaneously receive the reflected electromagnetic waves, and perform signal preprocessing and signal decomposition on the electromagnetic waves to obtain I / Q two-channel quadrature echo signals;

[0047] A signal processing module, configured to perform signal preprocessing on the I / Q two-channel quadrature echo signals to obtain processed signals;

[0048] A heart rate calculation module, configured to input the processed signals into a heart rate feature algorithm model to obtain a final heart rate result; the heart rate feature algorithm model includes a heart rate feature extraction sub-model and a heart rate acquisition sub-model; the heart rate acquisition sub-model is a regression model based on a generative adversarial network.

[0049] The BCG heart rate detection device based on terahertz radar in this embodiment adopts a multi-channel terahertz radar module to simultaneously transmit electromagnetic waves to multiple targets to be measured through a MIMO antenna array, and at the same time uses a heart rate acquisition sub-model based on a generative adversarial network, realizing real-time heart rate detection of multiple targets while improving the accuracy of heart rate detection, thereby improving the reliability and practicality of the detection results.

[0050] Embodiment 2

[0051] The BCG heart rate detection device based on terahertz radar in this embodiment includes:

[0052] Multi-channel terahertz radar module, which is used to simultaneously transmit electromagnetic waves to multiple targets to be measured through a MIMO antenna array; at the same time, receive the reflected electromagnetic waves, and perform signal preprocessing and signal decomposition on the electromagnetic waves to obtain I / Q two-channel quadrature echo signals;

[0053] Signal processing module, which is used to perform signal preprocessing on the I / Q two-channel quadrature echo signals to obtain processed signals;

[0054] Heart rate calculation module, which is used to input the processed signals into the heart rate feature algorithm model to obtain the final heart rate result; the heart rate feature algorithm model includes a heart rate feature extraction sub-model and a heart rate acquisition sub-model; the heart rate acquisition sub-model is a regression model based on a generative adversarial network.

[0055] Specifically, the multi-channel terahertz radar module includes:

[0056] Transmission module and reception module;

[0057] The transmission module is used to mix the baseband signal and the local oscillator signal to the terahertz band and transmit it through the MIMO antenna array;

[0058] The reception module includes a beamforming unit, a down-conversion processing unit and a decomposition unit. The beamforming unit is used to separate multiple targets from the received reflected signals, and the decomposition unit is used to decompose the down-converted baseband signal into I / Q two-channel quadrature echo signals.

[0059] The MIMO antenna array is an 8×8 planar array, the adjacent antenna spacing is 0.8 - 1.2 times the operating wavelength, and the operating frequency band is 0.3 - 0.5 THz.

[0060] In a specific implementation process, the transmission module emits terahertz waves (the frequency range is usually between 0.1 Hz and 10 THz) to irradiate the target, and the reception module receives the reflected signals. Such signals mainly contain the following information:

[0061] Micro-movements and vibrations of the target: such as the minute displacements caused by the heartbeat, breathing, etc. of the human body;

[0062] Environmental reflections and echoes: including the reflected signals of walls, particulate matter in the air, and other objects;

[0063] Noise, including the electronic noise inside the device and the electromagnetic interference from the external environment;

[0064] Therefore, the signals received by the reception module are a kind of modulated electromagnetic wave signals, which contain the dynamic information of the target and environmental noise.

[0065] Furthermore, the signal flow in the multi-channel terahertz radar module can be understood through the following steps:

[0066] (1) Generation and Modulation of Signals

[0067] Baseband Signal: The minute movements of the target (such as heartbeat, breathing) generate low-frequency baseband signals.

[0068] Local Oscillator Signal: A stable high-frequency reference signal is generated by a phase-locked loop (PLL) or other oscillators.

[0069] RF Up-Conversion: The baseband signal is mixed with the local oscillator signal through a mixer, and the frequency of the signal is increased to the terahertz band. This process is called up-conversion, and the generated high-frequency signal is the transmitted electromagnetic wave signal.

[0070] (2) Transmission and Reception of Signals

[0071] Transmitted Signal: The electromagnetic wave signal after up-conversion is transmitted through the transmission module to irradiate the target.

[0072] Reflected Signal: After the electromagnetic wave signal irradiates the target, the reflected signal contains the dynamic information of the target (such as minute movements) and environmental noise.

[0073] (3) Reception and Processing of Signals

[0074] Received Signal: The receiving module captures the reflected electromagnetic wave signal.

[0075] Down-Conversion: The received high-frequency signal is mixed with the local oscillator signal through a mixer, and the frequency of the signal is converted from the terahertz band back to the low-frequency signal (baseband signal).

[0076] Signal Processing: Multiple target separation is performed on the baseband signal after down-conversion, and at the same time, the baseband signal after down-conversion is decomposed into two orthogonal echo signals of I / Q.

[0077] Among them, the I-channel orthogonal signal is the component in phase with the local oscillator signal, and the Q-channel orthogonal signal is the component with a 90° phase difference from the local oscillator signal.

[0078] The two orthogonal echo signals of I / Q contain the complete information (amplitude and phase) of the target reflected signal.

[0079] The electromagnetic wave signal is a high-frequency signal used for transmission and reception in the radar system, which contains the dynamic information of the target and environmental noise.

[0080] The baseband signal is the original low-frequency signal, which contains the actual information of the target (such as heart rate signal), and needs to be modulated onto the high-frequency signal for transmission.

[0081] The local oscillator signal is the reference signal for mixing operations, ensuring signal synchronization and frequency conversion.

[0082] In this embodiment, the signal processing module includes:

[0083] A denoising unit that filters the I / Q quadrature echo signals in the 0.5 - 4 Hz frequency band using an adaptive band - pass filter to obtain the first I / Q quadrature echo signals;

[0084] A wavelet transform unit that performs 3 - layer decomposition using the Daubechies 4 wavelet basis, retains the coefficients in the 0.5 - 4 Hz frequency band, and obtains the second I / Q quadrature echo signals;

[0085] A correction unit that performs 5 - layer mode separation on the respiratory harmonic interference through variational mode decomposition, removes the first 3 layers of low - frequency respiratory components, and obtains the I / Q quadrature echo signals with respiratory harmonic interference removed as the processed signals.

[0086] In the correction unit, the number of modes K for variational mode decomposition is 6, the penalty factor α is 2000, and the iteration stop condition is that the modal energy change rate ≤ 1%.

[0087] In the specific implementation process, the signal is decomposed into K intrinsic mode functions through variational mode decomposition, and the last three layers (corresponding to the frequency 2 - 4 Hz) are selected as the heart - rate correlation, filtering out the first 3 layers of low - frequency respiratory harmonic components (frequency < 1 Hz);

[0088] When the modal energy change rate between two adjacent iterations ≤ 1%, the decomposition is terminated. The formula for the modal energy change rate is:

[0089]

[0090] where, is the energy of the k - th mode in the n - th iteration, is the energy of the k - th mode in the (n - 1) - th iteration, and ||.||2 represents the L2 norm;

[0091] When the modal energy change rate between two adjacent iterations ≤ 1%, the screened intrinsic mode function components are superimposed to obtain the I / Q quadrature echo signals with respiratory harmonic interference removed as the processed signals.

[0092] In the specific implementation process, the heart - rate feature algorithm model includes:

[0093] A heart - rate feature extraction sub - model that extracts the time - frequency features of the signal through multi - scale wavelet decomposition and dilated convolution;

[0094] A heart - rate acquisition sub - model that inputs the time - frequency features into the pre - constructed heart - rate acquisition sub - model to obtain the final heart - rate result.

[0095] Specifically, the heart - rate feature extraction sub - model includes:

[0096] An input layer that inputs the processed signal;

[0097] Wavelet transform decomposition layer, which is used to perform 4-layer decomposition using Morlet wavelet basis, covering the frequency band of 0.5 - 4 Hz, and obtaining signals of different scales;

[0098] The mathematical expression of the Morlet wavelet basis function is:

[0099]

[0100] where, w0 is the central angular frequency, is the Gaussian window function, and j is the imaginary unit.

[0101] The signal covers the frequency band of 0.5 - 4 Hz through 4-layer decomposition. The scale parameters of each layer are s = {8, 16, 32, 64}, and the corresponding frequency ranges are 3 - 4 Hz, 1.5 - 3 Hz, 0.75 - 1.5 Hz, and 0.5 - 0.75 Hz respectively.

[0102] Convolutional neural network layer, which is used to perform feature extraction on the signals at each scale, and capture the local and global features of the signals; the convolutional neural network layer uses a one-dimensional convolutional layer with a kernel size of 5 and a dilated convolutional layer with a dilation factor of 3;

[0103] Among them, the one-dimensional convolutional layer has a kernel size of 5, a stride of 1, an output channel number of 32, and uses ReLU activation; the dilated convolutional layer has a dilation factor of 3, a kernel size of 5, and an output channel number of 64. The convolutional neural network layer also uses a max pooling layer, and the pooling window of the max pooling layer is 3 and the stride is 2.

[0104] Feature fusion layer, which is used to obtain the attention weights of multi-scale different features through SENet and obtain a weighted feature vector;

[0105] Specifically, obtaining the attention weights of multi-scale features through SENet specifically includes:

[0106] Compression: Global average pooling along the time dimension;

[0107] Excitation: Learning the channel interdependence through a fully connected layer;

[0108] Weight weighting: Multiplying the learned weights with the meta feature map channel by channel to generate a weighted feature vector.

[0109] Output layer, which outputs the weighted heart rate features including heart rate variability and rhythm pattern.

[0110] In this embodiment, the heart rate acquisition sub-model includes a generator and a discriminator;

[0111] The generator is a 3-layer LSTM network with a hidden layer dimension of 128; the discriminator is a 4-layer one-dimensional convolutional network.

[0112] The heart rate acquisition sub - model is a sub - model trained with a training data set;

[0113] The training data set includes true heart rate signals of different historical environmental conditions and different target individuals, as well as corresponding heart rate variability and rhythm patterns.

[0114] When training the heart rate acquisition sub - model, the generator and the discriminator are alternately trained. The loss function of the generator is the weighted sum of the mean square error and the adversarial loss, and the loss function of the discriminator is the cross - entropy loss;

[0115] When the decrease rate of the generator loss function is less than 1% in 10 consecutive iterations, the training is terminated. The decrease rate is calculated by the moving window average method.

[0116] In the specific implementation process, training the heart rate acquisition sub - model includes:

[0117] Data acquisition: Collect true heart rate signals and corresponding heart rate variability and rhythm patterns, ensuring that the data covers different environmental conditions, different target individuals, and various possible interference situations.

[0118] Data annotation: Annotate the collected signals to distinguish true signals from noise / interference components. It can be completed with the assistance of expert annotation or automated algorithms.

[0119] Data pre - processing: Pre - process the collected signals, including operations such as filtering, denoising, and normalization, to improve data quality and reduce noise interference during training.

[0120] Data augmentation: Expand the data set through data augmentation techniques (such as adding noise, adjusting signal amplitude, time shift, etc.) to enhance the generalization ability of the model.

[0121] Model initialization: Randomly initialize the weights of the generator to ensure that its initial state can generate signals with a certain degree of randomness. Common initialization methods, such as Xavier initialization or He initialization, can be used. Similarly, randomly initialize the weights of the discriminator so that its ability to distinguish true signals from generated signals is close to random guessing at the beginning of training.

[0122] The adversarial training process includes generator training and discriminator training:

[0123] Generator training: The generator starts from a random noise vector (usually following a normal distribution) or an initial hidden state and generates a series of fake heart rate signal samples. The generated fake samples are input into the discriminator, and the discriminator outputs the probability that these fake samples are true signals; Calculate the gradient of the generator through backpropagation and use an optimization algorithm (such as Adam or RMSprop) to update the parameters of the generator to minimize the loss function of the generator;

[0124] Discriminator Training: Randomly extract real heart rate signal samples from the dataset and input them into the discriminator. Input the fake samples generated by the generator into the discriminator. Calculate the gradient of the discriminator through backpropagation and use an optimization algorithm to update the parameters of the discriminator to minimize the loss function of the discriminator.

[0125] In each training iteration, alternately train the generator and the discriminator. Usually, train the discriminator several times first, and then train the generator once. This alternating training method helps to maintain the balance between the generator and the discriminator.

[0126] Furthermore, dynamically adjust the training steps of the generator and the discriminator according to the loss curve and the quality of the generated samples during the training process. For example, if the loss of the discriminator is too low, it may be necessary to increase the training steps of the generator; vice versa.

[0127] A BCG heart rate detection device based on terahertz radar in this embodiment improves the accuracy of heart rate detection while realizing real-time heart rate detection of multiple targets, thereby improving the reliability and practicality of the detection results.

[0128] Embodiment 3

[0129] A BCG heart rate detection device based on terahertz radar in this embodiment is deployed in a remote monitoring ward of a certain hospital to perform real-time heart rate monitoring on multiple patients. Among them, the multi-channel terahertz radar module is set on the ceiling of the ward to ensure that the radar wave can cover the entire ward area. The radar operating frequency is set between 0.1 THz and 10 THz to ensure sufficient penetration and resolution.

[0130] Install the signal processing module and the heart rate calculation module in the console of the ward and connect them to the multi-channel terahertz radar module.

[0131] Locate the positions of multiple patients in the ward through the scanning function of the terahertz radar. The radar device can identify the minute movements of the human body (such as chest vibrations caused by breathing and heartbeat).

[0132] Then a BCG heart rate detection device based on terahertz radar in this embodiment includes:

[0133] A multi-channel terahertz radar module for simultaneously transmitting electromagnetic waves to multiple targets to be measured through a MIMO antenna array; simultaneously receiving the reflected electromagnetic waves, and performing signal preprocessing and signal decomposition on the electromagnetic waves to obtain I / Q two-channel quadrature echo signals;

[0134] A signal processing module for performing signal preprocessing on the I / Q two-channel quadrature echo signals to obtain the processed signals;

[0135] A heart rate calculation module, which is used to input the processed signal into a heart rate feature algorithm model to obtain the final heart rate result; the heart rate feature algorithm model includes a heart rate feature extraction sub-model and a heart rate acquisition sub-model; the heart rate acquisition sub-model is a regression model based on a generative adversarial network.

[0136] At the same time, on the user interface of the nurse station, the heart rate monitoring results of each patient are displayed in real time, including information such as heart rate curves and heart rate variability.

[0137] Furthermore, the detection data is transmitted to the cloud server in real time for long-term storage and analysis. Medical staff can access the detection data through mobile devices or remote terminals to achieve remote monitoring and diagnosis.

[0138] A BCG heart rate detection device based on terahertz radar in this embodiment further improves the accuracy and efficiency of heart rate monitoring, and at the same time enables accurate monitoring of the heart rate of each target in a complex environment.

[0139] In the description of the present invention, it should be understood that the terms "first" and "second" are only used for descriptive purposes and cannot be construed as indicating or implying relative importance or implicitly specifying the quantity of the indicated technical features. Thus, the features defined with "first" and "second" may explicitly or implicitly include one or more of such features. In the description of the present invention, "a plurality of" means two or more unless otherwise specifically defined.

[0140] In the present invention, unless otherwise clearly specified and limited, the terms "installed", "connected", "connected", "fixed", etc. should be understood in a broad sense. For example, it can be a fixed connection, a detachable connection, or integrated; it can be a mechanical connection or an electrical connection; it can be directly connected or indirectly connected through an intermediate medium; it can be the internal connection of two components or the interaction relationship between two components. For those of ordinary skill in the art, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.

[0141] In the present invention, unless otherwise clearly specified and limited, the first feature being "on" or "under" the second feature can be that the first and second features are in direct contact, or the first and second features are indirectly in contact through an intermediate medium. Moreover, the first feature being "above", "over" and "on" the second feature can be that the first feature is directly above or obliquely above the second feature, or merely indicates that the first feature has a higher horizontal height than the second feature. The first feature being "under", "beneath" and "under" the second feature can be that the first feature is directly below or obliquely below the second feature, or merely indicates that the first feature has a lower horizontal height than the second feature.

[0142] In the description of this specification, the description of terms such as "one embodiment", "some embodiments", "embodiment", "example", "specific example" or "some examples" means that the specific features, structures, materials or characteristics described in connection with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic representations of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described may be combined in any one or more embodiments or examples in a suitable manner. In addition, without contradiction, those skilled in the art can combine and combine the different embodiments or examples described in this specification and the features of different embodiments or examples.

[0143] Although the embodiments of the present invention have been shown and described above, it can be understood that the above embodiments are exemplary and should not be construed as limiting the present invention. Those of ordinary skill in the art can make changes, modifications, substitutions and variations to the above embodiments within the scope of the present invention.

Claims

1. A BCG heart rate detection device based on terahertz radar, characterized in that, Including: A multi-channel terahertz radar module for simultaneously transmitting electromagnetic waves to multiple targets to be measured through a MIMO antenna array; simultaneously receiving the reflected electromagnetic waves, and performing signal preprocessing and signal decomposition on the electromagnetic waves to obtain I / Q two-channel quadrature echo signals; A signal processing module for performing signal preprocessing on the I / Q two-channel quadrature echo signals to obtain processed signals; A heart rate calculation module for inputting the processed signals into a heart rate feature algorithm model to obtain the final heart rate result; The heart rate feature algorithm model includes a heart rate feature extraction sub-model and a heart rate acquisition sub-model; the heart rate acquisition sub-model is a regression model based on a generative adversarial network.

2. The BCG heart rate detection device based on terahertz radar according to claim 1, wherein The multi-channel terahertz radar module includes: A transmitting module and a receiving module; The transmitting module is used to mix the baseband signal with the local oscillator signal to the terahertz band and transmit it through the MIMO antenna array; The receiving module includes a beamforming unit, a down-conversion processing unit, and a decomposition unit. The beamforming unit is used to separate multiple targets from the received reflected signals, and the decomposition unit is used to decompose the down-converted baseband signal into I / Q two-channel quadrature echo signals.

3. The terahertz radar-based BCG heart rate detection device according to claim 2, wherein The MIMO antenna array is an 8×8 planar array, the adjacent antenna spacing is 0.8-1.2 times the operating wavelength, and the operating frequency band is 0.3-0.5 THz.

4. The BCG heart rate detection device based on terahertz radar according to claim 1, characterized in that, The signal processing module includes: A denoising unit that filters the I / Q two-channel quadrature echo signals in the 0.5-4 Hz frequency band using an adaptive band-pass filter to obtain the first I / Q two-channel quadrature echo signals; A wavelet transform unit that performs 3-layer decomposition using Daubechies 4 wavelet basis and retains the coefficients in the 0.5-4 Hz frequency band to obtain the second I / Q two-channel quadrature echo signals; A correction unit for performing 5-layer modal separation on the respiratory harmonic interference through variational mode decomposition, removing the first 3 layers of low-frequency respiratory components, and obtaining the I / Q two-channel quadrature echo signals without respiratory harmonic interference as the processed signals.

5. The BCG heart rate detection device based on terahertz radar according to claim 1, wherein, The heart rate feature algorithm model includes: A heart rate feature extraction sub-model for extracting the time-frequency features of the signals through multi-scale wavelet decomposition and dilated convolution; A heart rate acquisition sub-model for inputting the time-frequency features into a pre-constructed heart rate acquisition sub-model to obtain the final heart rate result.

6. The BCG heart rate detection device based on terahertz radar according to claim 5, characterized in that The heart rate feature extraction sub-model includes: An input layer for inputting the processed signals; A wavelet transform decomposition layer for performing 4-layer decomposition using Morlet wavelet basis, covering the 0.5-4 Hz frequency band, to obtain signals of different scales; A convolutional neural network layer for performing feature extraction on the signals at each scale to capture the local and global features of the signals; the convolutional neural network layer uses a one-dimensional convolutional layer with a kernel size of 5 and a dilated convolutional layer with a dilation factor of 3; A feature fusion layer for obtaining the attention weights of multi-scale different features through SENet to obtain a weighted feature vector; An output layer for outputting the weighted heart rate features including heart rate variability and rhythm pattern.

7. The BCG heart rate detection device based on terahertz radar according to claim 5, wherein The heart rate acquisition sub-model includes a generator and a discriminator; The generator is a 3-layer LSTM network with a hidden layer dimension of 128; The discriminator is a 4-layer one-dimensional convolutional network.

8. The BCG heart rate detection device based on a terahertz radar according to claim 7, characterized in that, The heart rate acquisition sub-model is a sub-model trained using a training data set; The training data set includes true heart rate signals, corresponding heart rate variability, and rhythm patterns of different historical environmental conditions and different target individuals.

9. The terahertz radar-based BCG heart rate detection device according to claim 8, wherein When training the heart rate acquisition sub-model, the generator and the discriminator are alternately trained. The loss function of the generator is the weighted sum of the mean square error and the adversarial loss, and the loss function of the discriminator is the cross-entropy loss; When the descent rate of the generator loss function is less than 1% in 10 consecutive iterations, the training is terminated, and the descent rate is calculated by the moving window averaging method.

10. The BCG heart rate detection device based on terahertz radar according to claim 4, wherein In the correction unit, the number of modes K of the variational mode decomposition is 6, the penalty factor α is 2000, and the iteration stop condition is that the modal energy change rate ≤ 1%.

Images