Radar blood pressure non-contact detection method based on CNN-BiLSTM structure and related device

Abstract

The invention discloses a radar blood pressure non-contact detection method based on a CNN-BiLSTM structure and a related device. The radar blood pressure non-contact detection method comprises the steps that radial artery pulse wave signals are obtained; the radial artery pulse wave signal is input into the trained CNN-BiLSTM model, a blood pressure signal is obtained through prediction, radar blood pressure non-contact detection based on the CNN-BiLSTM structure is completed, non-contact blood pressure detection can be achieved through the method and the related device, and accuracy and universality are high.

Description

Technical Field

[0001] The present invention belongs to the field of non-contact detection technology and relates to a radar blood pressure non-contact detection method based on a CNN-BiLSTM structure and a related device. Background Art

[0002] In the field of modern healthcare monitoring, blood pressure measurement, as a key tool for assessing cardiovascular health, has long been a hot topic in research and application. Traditional blood pressure measurement techniques primarily rely on contact devices, such as arterial cannulation, Korotkoff sound auscultation, oscillometric methods, volume compensation methods, arterial tension methods, ultrasonic pressure measurement, and blood pressure measurement based on photoplethysmography (PPG). While widely used in clinical practice, these methods generally have limitations. For example, arterial cannulation is invasive, causing pain and potential risks for patients. Korotkoff sound auscultation, while currently the gold standard for blood pressure measurement, requires specialized personnel and is a cumbersome process. Finally, blood pressure measurement based on photoplethysmography requires at least two sensors to simultaneously measure two signals to obtain pulse wave transit time, which increases the burden on the subject and requires strict synchronization of the two sensors, making measurement more challenging.

[0003] Furthermore, contact blood pressure measurement methods face other challenges in practical applications. For example, measurement requires the sensor to be in contact with the human body, and even requires applying a certain amount of pressure to the blood vessel walls. This can cause venous congestion at the measurement site, thereby affecting measurement accuracy. Furthermore, contact measurement methods place high demands on sensor placement and detection accuracy, making them unsuitable for certain special populations (such as patients with extensive burns, skin diseases, those allergic to the cuff or sensor, and newborn infants), limiting their scope of application. Furthermore, contact blood pressure measurement methods can also produce the "white coat effect," whereby a patient's nervousness during blood pressure measurement in a hospital or clinic can cause their blood pressure to rise, affecting the accuracy of the measurement results.

[0004] In recent years, with the rapid development of bioradar technology, non-contact blood pressure measurement has become a hot topic of research. Bioradar technology exploits the interaction between electromagnetic waves and biological tissues. By emitting electromagnetic waves of a specific frequency and receiving the reflected or transmitted signals, it can detect and identify minute movements of living organisms. This technology, without requiring any sensors or electrodes, can detect minute movements of the human skin surface at a distance and without contact. It offers advantages such as being contactless, unconstrained, unaffected by light, penetrating certain areas (such as clothing), and capable of continuous measurement. However, current bioradar-based non-contact blood pressure measurement technology still faces several challenges. First, pulse wave signals are weak, non-linear, and non-stationary signals, making them susceptible to interference from high-frequency noise and low-frequency drift during acquisition. These noise and interference can affect the feature extraction and processing of pulse wave signals, thereby reducing the accuracy of blood pressure measurement. Second, existing bioradar-based blood pressure measurement methods mostly estimate blood pressure for a single individual and fail to account for individual differences, such as differences in blood density, arterial wall thickness, and vessel length. This results in insufficient generalizability of the blood pressure measurement models. Furthermore, while some machine learning-based blood pressure detection methods can extract features from pulse wave signals, their feature extraction process relies on manual definition and calculation. Some features are vaguely defined and difficult to quantify, which can affect the accuracy of experimental results. Furthermore, these methods have limited modeling capabilities when processing complex data, making it difficult to effectively capture high-dimensional, complex features, thus limiting further improvements in blood pressure detection accuracy.

[0005] In summary, existing blood pressure monitoring technologies still have significant flaws and shortcomings in terms of contact measurement limitations, non-contact measurement signal processing accuracy, and adaptability to individual differences. These issues not only affect the accuracy and reliability of blood pressure monitoring but also restrict its widespread adoption and application across a wider range of populations and scenarios. Therefore, there is an urgent need for a new non-contact blood pressure monitoring technology that overcomes the shortcomings of existing technologies, achieves high accuracy, and is widely applicable to meet the needs of clinical and home health management. Summary of the Invention

[0006] The purpose of the present invention is to overcome the shortcomings of the above-mentioned prior art and provide a radar blood pressure non-contact detection method and related devices based on the CNN-BiLSTM structure. This method and related devices can realize non-contact blood pressure detection with high accuracy and universality.

[0007] To achieve the above objectives, the present invention discloses a radar blood pressure non-contact detection method based on a CNN-BiLSTM structure, comprising:

[0008] Acquire radial artery pulse wave signal;

[0009] The radial artery pulse wave signal is input into the trained CNN-BiLSTM model to predict the blood pressure signal, thereby completing the radar blood pressure non-contact detection based on the CNN-BiLSTM structure.

[0010] The further improvement of the radar non-contact blood pressure detection method based on the CNN-BiLSTM structure of the present invention is:

[0011] Furthermore, the process of obtaining the radial artery pulse wave signal is as follows:

[0012] A 24 GHz continuous wave radar was used to collect radial artery pulse wave signals.

[0013] Furthermore, the step of inputting the radial artery pulse wave signal into the trained CNN-BiLSTM model to predict the blood pressure signal further includes:

[0014] Build a training dataset;

[0015] Build a CNN-BiLSTM model;

[0016] The CNN-BiLSTM model is trained and tested based on the training data set to obtain a trained CNN-BiLSTM model.

[0017] Furthermore, the process of constructing the training data set is:

[0018] The training dataset is constructed based on the fixed-length sliding window sample segmentation method.

[0019] Furthermore, the CNN-BiLSTM model includes a CNN network, a bidirectional long short-term memory network, a fully connected layer and an output layer, wherein the CNN network, the bidirectional long short-term memory network, the fully connected layer and the output layer are connected in sequence.

[0020] Furthermore, the process of training and testing the CNN-BiLSTM model based on the training data set to obtain the trained CNN-BiLSTM model is as follows:

[0021] The CNN-BiLSTM model is trained and tested using Adam based on batch normalization and dropout technology to obtain a trained CNN-BiLSTM model.

[0022] The present invention discloses a radar blood pressure non-contact detection system based on a CNN-BiLSTM structure, comprising:

[0023] An acquisition module, used for acquiring radial artery pulse wave signals;

[0024] The prediction module is used to input the radial artery pulse wave signal into the trained CNN-BiLSTM model to predict the blood pressure signal and complete the radar blood pressure non-contact detection based on the CNN-BiLSTM structure.

[0025] The radar blood pressure contactless detection system based on the CNN-BiLSTM structure of the present invention is further improved in that:

[0026] Furthermore, it also includes:

[0027] The first building module is used to build a training data set;

[0028] The second building block is used to build the CNN-BiLSTM model;

[0029] The training module is used to train and test the CNN-BiLSTM model based on the training data set to obtain a trained CNN-BiLSTM model.

[0030] The present invention discloses a computer device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the steps of the radar blood pressure non-contact detection method based on a CNN-BiLSTM structure are implemented.

[0031] The present invention discloses a computer-readable storage medium, which stores a computer program. When the computer program is executed by a processor, the steps of the radar blood pressure non-contact detection method based on the CNN-BiLSTM structure are implemented.

[0032] The present invention has the following beneficial effects:

[0033] During specific operation, the radar-based non-contact blood pressure detection method and related device based on the CNN-BiLSTM structure described in the present invention inputs the radial artery pulse wave signal into the trained CNN-BiLSTM model to predict the blood pressure signal. The present invention, through the combination of CNN and BiLSTM, can not only extract the local features of the pulse wave signal, but also capture its temporal dependence, further improving the performance and robustness of the model. CNN is responsible for processing the spatial information and local patterns in the pulse wave signal, while BiLSTM, through its internal memory units and gating mechanism, can effectively process the temporal information in the feature matrix and discover potential correlations between different features, thereby further improving the accuracy of blood pressure detection. In this way, the entire model can not only extract the key features of the pulse wave signal, but also understand its temporal variation patterns, thereby providing more accurate and reliable results for blood pressure detection. BRIEF DESCRIPTION OF THE DRAWINGS

[0034] The accompanying drawings, which constitute part of the present invention, are intended to provide a further understanding of the present invention. The exemplary embodiments of the present invention and their descriptions are intended to explain the present invention and do not constitute an undue limitation of the present invention. In the accompanying drawings:

[0035] Figure 1 Schematic diagram of the specific process of sliding window in the present invention;

[0036] Figure 2 The structure diagram of the CNN-BiLSTM model;

[0037] Figure 3a This is the effect diagram of the sigmoid activation function;

[0038] Figure 3b This is the effect diagram of the Tanh activation function;

[0039] Figure 3c This is the effect diagram of the ReLU activation function;

[0040] Figure 4 This is the structural diagram of LSTM neurons;

[0041] Figure 5 This is the structural diagram of the BiLSTM network;

[0042] Figure 6a This is the systolic blood pressure detection result diagram of CNN-LSTM;

[0043] Figure 6b This is the diastolic blood pressure detection result diagram of CNN-LSTM;

[0044] Figure 7a This is the systolic blood pressure detection result diagram of CNN-BiLSTM;

[0045] Figure 7b This is the diastolic blood pressure detection result diagram of CNN-BiLSTM;

[0046] Figure 8a Bland-Altman plot of systolic and diastolic blood pressure;

[0047] Figure 8b Bland-Altman plot of diastolic blood pressure. DETAILED DESCRIPTION

[0048] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of them. All other embodiments obtained by ordinary technicians in this field based on the embodiments of the present invention without making any creative efforts shall fall within the scope of protection of the present invention.

[0049] In the description of the present invention, it is to be understood that the terms “include” and “comprise” indicate the presence of the described features, wholes, steps, operations, elements and / or components, but do not exclude the presence or addition of one or more other features, wholes, steps, operations, elements, components and / or collections thereof.

[0050] It should also be understood that the terms used in the present specification are only for the purpose of describing particular embodiments and are not intended to limit the present invention. As used in the present specification and the appended claims, the singular forms "a", "an", and "the" are intended to include the plural forms unless the context clearly indicates otherwise.

[0051] It should be further understood that the term "and / or" as used in the present specification and the appended claims refers to and includes any and all possible combinations of one or more of the associated listed items. For example, A and / or B may represent: A exists alone, A and B exist simultaneously, and B exists alone. In addition, the character " / " in the present invention generally indicates that the associated objects are in an "or" relationship.

[0052] It should be understood that although the terms "first," "second," and "third" may be used to describe preset ranges in embodiments of the present invention, these preset ranges should not be limited to these terms. These terms are merely used to distinguish one preset range from another. For example, without departing from the scope of embodiments of the present invention, the first preset range may also be referred to as the second preset range, and similarly, the second preset range may also be referred to as the first preset range.

[0053] The word "if," as used herein, may be interpreted as "at the time of" or "when" or "in response to determining" or "in response to detecting," depending on the context. Similarly, the phrases "if it is determined" or "if (stated condition or event) is detected" may be interpreted as "when it is determined" or "in response to the determination" or "when detecting (stated condition or event)" or "in response to detecting (stated condition or event)," depending on the context.

[0054] In order to make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Generally, the components of the embodiments of the present invention described and shown in the drawings herein can be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of the present invention provided in the drawings is not intended to limit the scope of the claimed invention, but merely represents selected embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0055] The accompanying drawings illustrate various schematic diagrams of structures according to embodiments disclosed herein. These figures are not drawn to scale; for clarity, some details are exaggerated and some details may be omitted. The shapes of the various regions and layers shown in the figures, as well as their relative sizes and positional relationships, are merely exemplary and may deviate in practice due to manufacturing tolerances or technical limitations. Those skilled in the art may design regions / layers with different shapes, sizes, and relative positions as needed.

[0056] Example 1

[0057] The radar blood pressure non-contact detection method based on the CNN-BiLSTM structure of the present invention comprises the following steps:

[0058] 1) Obtain radial artery pulse wave signal;

[0059] It should be noted that the present invention uses a 24 GHz continuous wave radar to collect radial artery pulse wave signals.

[0060] 2) Inputting the radial artery pulse wave signal into the trained CNN-BiLSTM model to obtain a blood pressure signal.

[0061] The specific process of step 2) is:

[0062] 21) Constructing a training dataset;

[0063] A 24GHz continuous wave radar is used to collect radial artery pulse wave signals, and an Omron electronic sphygmomanometer is used to measure the reference blood pressure value. The collected radial artery pulse wave signals are preprocessed to obtain pure radial artery pulse wave signals, and then a training dataset is constructed based on the fixed-length sliding window sample segmentation method.

[0064] refer to Figure 1 ,The process of constructing a training data set based on the fixed-length sliding window sample segmentation method is:

[0065] 1a) Identify and record the starting point and end point of each radar pulse wave signal, and record the position of each pulse wave.

[0066] 2a) Determine the size L of the sliding window. To ensure that there is sufficient pulse wave signal in each window, in this embodiment, L is 5s.

[0067] 3a) Starting from the starting point, the radar pulse wave signal is intercepted at each sample, and the window slides backward in steps of 2 seconds, intercepting consecutive segments in sequence until the signal ends.

[0068] Each collected signal segment is divided into 38 independent samples through the fixed-length sliding window sample segmentation method. Each independent sample corresponds to the blood pressure value measured by the Omron electronic blood pressure monitor as a reference. All samples processed in this way are used as training data for the model, and the training data set is constructed based on this.

[0069] 22) Build a CNN-BiLSTM model;

[0070] The CNN-BiLSTM model includes a convolutional neural network (CNN), a bidirectional long short-term memory network (BiLSTM), a fully connected layer and an output layer, wherein the CNN network, the bidirectional long short-term memory network, the fully connected layer and the output layer are connected in sequence.

[0071] like Figure 2 As shown in the figure, the detection process of the CNN-BiLSTM model is as follows: the cut time series is used as input vectors. These input vectors completely retain the timing information of the pulse wave signal. Subsequently, the input vectors are sent to the CNN for deep feature extraction, and a feature matrix rich in key information is generated from the input vectors. After feature extraction by CNN, the generated feature matrix is ​​input into the two-layer BiLSTM network for association detection. Finally, the detection results are passed to the output layer through the fully connected layer, thereby achieving accurate blood pressure detection.

[0072] Specifically, for the CNN layer, although the convolution operation is mainly used to process two-dimensional images, its principle can also be used to model one-dimensional time series data, and the calculation method is roughly the same as the two-dimensional case. Therefore, the convolution kernel needs to be adapted according to the dimension of the input data. The present invention targets the one-dimensional pulse wave signal processing task and constructs a one-dimensional convolutional neural network structure that can effectively extract its features. The input data is a pulse wave signal sequence divided by time intervals. Let the input sequence be X r ,r=1,2,…m, the convolution kernel size is set to w t,t=1,2,…,p, the feature map output by the previous layer is calculated with the convolution kernel in sequence, and the bias term b is set. The convolution output is:

[0073]

[0074] In deep neural networks, the connections between layers are usually modeled by a nonlinear function, which is called an activation function. To improve model training efficiency and alleviate the gradient vanishing problem, this model introduces ReLU as an activation function after the convolutional layer output. Represents the activation function. Figure 3a 、 Figure 3b and Figure 3c As shown in Figure 2, the comparative effects of various common activation functions in deep learning are demonstrated.

[0075] In the field of deep learning, the gradients of Sigmoid and tanh activation functions are close to zero in the saturation region, which can easily cause gradient vanishing, especially in deep networks, which leads to a decrease in the training convergence speed and limits the further deepening of the network. In contrast, the ReLU function can effectively alleviate the gradient vanishing problem of deep networks because it has a constant gradient in most input ranges. In addition, the asymmetric activation mechanism of ReLU is more in line with the response characteristics of biological neurons, which helps to improve the training efficiency of the model. Therefore, after comprehensive comparison, the present invention finally adopts the ReLU function as the activation unit of the CNN network.

[0076] For the BiLSTM layer, the bidirectional long short-term memory network is an improved structure of the recurrent neural network. By introducing memory units that can save historical states, it improves the feature modeling ability of long time series data. LSTM relies on a unique gating mechanism to achieve dynamic adjustment of information retention and discarding, thereby controlling the update process of cell states. The long short-term memory network module contains three core gating units, namely the forget gate f t , input gate i t and output gate o t , its internal structure is as follows Figure 4 shown.

[0077] In LSTM, the forget gate f t Controls the degree of retention of cell state information in the unit at the previous moment, input gate i t Combined with the current input features to optimize the update strength of the cell state, the output gate o t Based on the current cell state and hidden state, the final time series output signal is generated. The cell state C t With the hidden state h t The update process is:

[0078]

[0079] Among them, σ represents the sigmoid activation function, which is responsible for the feature mapping process of the gate unit; h t-1 Indicates the hidden state of the previous moment; x t Input vector for the current moment; b x Represents the bias term of the corresponding gate.

[0080] BiLSTM is an enhanced structure of LSTM, which consists of two LSTM networks, forward and backward, to achieve global perception of the context features of the input sequence and effectively extract the forward and reverse dependencies of the sequence data. Its network structure is as follows: Figure 5 shown.

[0081] The hidden layer state O of BiLSTM at time t t From the forward hidden layer state and the backward hidden layer state The common components are:

[0082]

[0083] Among them, f represents the operation function inside the LSTM unit, g is the activation function ReLU, and w x represents the weight parameter of the associated input.

[0084] As for the fully connected layer, it is a common structural form in neural networks. In this layer, each neuron is interconnected with the neurons in the previous layer and is assigned corresponding weight parameters. This type of structure is generally located between the hidden layer and the output layer and is one of the important modules of deep neural networks.

[0085] 23) Using the training dataset to train and test the CNN-BiLSTM model.

[0086] The model constructed in the present invention mainly includes an input layer, two convolutional layers, two pooling layers, two BiLSTM layers, a fully connected layer, and an output layer. The loss function uses the mean square error (MSE), and the optimizer uses Adam, which has strong adaptive learning capabilities, to improve the convergence speed of model training. The initial learning rate is set to 0.001, and the learning rate decay factor is set to 0.9 to control the amplitude of weight updates during the model iteration process. The number of training rounds is set to 200, indicating that the entire training data will be fully iterated 200 times. The batch size is 32, that is, 32 samples are used in each training round to participate in the gradient update simultaneously. This parameter directly affects the training efficiency and optimization level of the model. The parameters of each module are shown in Table 1.

[0087] Table 1

[0088]

[0089] Note: “~” means no parameter

[0090] To enhance the model's robustness and generalization, the pulse wave signals extracted by convolution are batch normalized. As an effective data preprocessing method, batch normalization dynamically adjusts the output distribution of each intermediate layer, ensuring that the inputs received by subsequent network layers remain relatively consistent. This consistency helps improve model training. On the one hand, it accelerates model convergence, enabling the model to achieve optimal performance within fewer iterations; on the other hand, it can somewhat suppress overfitting. Furthermore, because the BiLSTM model is prone to overfitting when the number of parameters is large, dropout technology was introduced for regularization. Dropout is a simple and effective regularization method. Its basic principle is to randomly "drop out" some neurons in the network during model training, even if these neurons do not participate in forward or backward propagation in the current iteration. This reduces the dependencies between neurons in the network and alleviates overfitting.

[0091] Confirmatory trials

[0092] According to the set fixed sliding window strategy, the radar echo sequence was segmented to obtain a total of 4020 groups of pulse wave signal samples. To verify the effectiveness of the proposed model, the three network architectures of CNN-BiLSTM, XgBoost, and CNN-LSTM described in the present invention were selected for comparison. Among them, the basic CNN model contains two convolutional layers and a pooling layer, as well as a fully connected layer structure. The preprocessed pulse wave signal is first input into the convolutional layer to extract primary features, and then the downsampling process is completed through the pooling layer operation, thereby reducing the feature dimension and retaining the core information. This feature extraction process is continuously iterated in the network, capturing higher-level abstract representations step by step. Finally, the features output at the end of the network will be passed to the flattening layer and the fully connected module to achieve feature integration and conversion, and output the blood pressure prediction value.

[0093] Due to the large amount of sample data, only 200 predicted blood pressure values ​​are selected for depiction for the convenience of observation. The systolic and diastolic blood pressure prediction results of CNN-LSTM are shown in the figure. Figure 6a and Figure 6b As shown in Figure 2, the systolic and diastolic blood pressure prediction results of CNN-BiLSTM are as follows: Figure 7a and Figure 7b As shown in the figure, it can be intuitively observed that the reference blood pressure value obtained by the Omron electronic blood pressure monitor has a good fit with the CNN-BiLSTM detection value.

[0094] The present invention compares and evaluates the performance of XgBoost, CNN-LSTM, and CNN-BiLSTM models by using indicators such as relative error, root mean square error, and determination coefficient. The results are shown in Table 2. Except for the MAE indicators of diastolic blood pressure and MAPE indicators of systolic blood pressure, the blood pressure detection model of the deep learning model CNN-BiLSTM is more accurate than that of CNN-LSTM and Xgboost for feature extraction and detection.

[0095] Table 2

[0096]

[0097]

[0098] The consistency analysis of the actual blood pressure value and the prediction result based on CNN-BiLSTM is as follows: Figure 8a and Figure 8b As shown in the figure, 95% of the data fall within the consistency interval of MD±1.96SD. In view of this, it shows that the blood pressure values ​​detected by the CNN-BiLSTM method are consistent with the blood pressure values ​​measured by the electronic sphygmomanometer.

[0099] Example 2

[0100] The radar blood pressure non-contact detection system based on the CNN-BiLSTM structure of the present invention includes:

[0101] An acquisition module, used for acquiring radial artery pulse wave signals;

[0102] The prediction module is used to input the radial artery pulse wave signal into the trained CNN-BiLSTM model to predict the blood pressure signal and complete the radar blood pressure non-contact detection based on the CNN-BiLSTM structure.

[0103] In this embodiment, it also includes:

[0104] The first building module is used to build a training data set;

[0105] The second building block is used to build the CNN-BiLSTM model;

[0106] The training module is used to train and test the CNN-BiLSTM model based on the training data set to obtain a trained CNN-BiLSTM model.

[0107] The division of modules in the embodiments of the present application is illustrative and is merely a logical functional division. In actual implementation, other division methods may be used. Furthermore, the functional modules in the various embodiments of the present application may be integrated into a single processor, or may exist physically separately, or two or more modules may be integrated into a single module. The aforementioned integrated modules may be implemented in the form of hardware or software functional modules.

[0108] Example 3

[0109] A computer device includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the steps of the radar-based non-contact blood pressure detection method based on a CNN-BiLSTM architecture are implemented, for example, including: obtaining a radial artery pulse wave signal; inputting the radial artery pulse wave signal into a trained CNN-BiLSTM model to predict a blood pressure signal, thereby completing the radar-based non-contact blood pressure detection based on the CNN-BiLSTM architecture. The memory may include internal memory, such as a high-speed random access memory (RAM), or non-volatile memory, such as at least one disk drive. The processor, network interface, and memory are interconnected via an internal bus, which may be an industrial standard architecture bus, a peripheral component interconnect standard bus, an extended industrial standard architecture bus, or the like. The bus may be classified as an address bus, a data bus, a control bus, or the like. The memory is used to store programs. Specifically, the programs may include program code, which includes computer operating instructions. The memory may include both internal memory and non-volatile memory, and provides instructions and data to the processor.

[0110] Example 4

[0111] A computer-readable storage medium stores a computer program. When the computer program is executed by a processor, it implements the steps of the radar blood pressure non-contact detection method based on the CNN-BiLSTM structure, for example, including: obtaining a radial artery pulse wave signal; inputting the radial artery pulse wave signal into a trained CNN-BiLSTM model to predict a blood pressure signal, thereby completing the radar blood pressure non-contact detection based on the CNN-BiLSTM structure. Specifically, the computer-readable storage medium includes, but is not limited to, volatile memory and / or non-volatile memory. The volatile memory may include random access memory (RAM) and / or cache memory, etc. The non-volatile memory may include read-only memory (ROM), hard disk, flash memory, optical disk, magnetic disk, etc.

[0112] Those skilled in the art will appreciate that the embodiments of the present application can be provided as methods, systems, or computer program products. Therefore, the present application can adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment in combination with software and hardware. Moreover, the present application can adopt the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to magnetic disk storage, CD-ROM, optical storage, etc.) that contain computer-usable program code.

[0113] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the steps in the process. Figure 1 a process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0114] These computer program instructions may also be stored in a computer readable memory that can direct a computer or other programmable data processing device to work in a specific manner, so that the instructions stored in the computer readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 a process or multiple processes and / or boxes Figure 1 The function specified in one or more boxes.

[0115] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operational steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing the instructions executed on the computer or other programmable device for implementing the process. Figure 1 a process or multiple processes and / or boxes Figure 1 A step that specifies a function in one or more boxes.

[0116] Those skilled in the art will readily identify other embodiments of the present invention after considering the specification and disclosure of the invention. This application is intended to cover any variations, uses, or adaptations of the present invention that follow the general principles of the invention and include common knowledge or customary techniques in the art not disclosed herein. The description and examples are to be considered as exemplary only, with the true scope and spirit of the invention being indicated by the following claims.

[0117] It should be understood that the present invention is not limited to the exact construction described above and shown in the drawings, and that various modifications and changes may be made without departing from the scope thereof. The scope of the present invention is limited only by the appended claims.

[0118] The above description is only a preferred embodiment of the present invention and does not limit the present invention in any way. Any simple modification, change and equivalent structural change made to the above embodiment based on the technical essence of the present invention shall still fall within the scope of protection of the technical solution of the present invention.

Claims

1. A radar blood pressure non-contact detection method based on CNN-BiLSTM structure, characterized in that: include: Acquire radial artery pulse wave signal; The radial artery pulse wave signal is input into the trained CNN-BiLSTM model to predict the blood pressure signal, thereby completing the radar blood pressure non-contact detection based on the CNN-BiLSTM structure.

2. The radar blood pressure non-contact detection method based on the CNN-BiLSTM structure according to claim 1 is characterized in that, The process of obtaining the radial artery pulse wave signal is as follows: A 24 GHz continuous wave radar was used to collect radial artery pulse wave signals.

3. The radar blood pressure non-contact detection method based on the CNN-BiLSTM structure according to claim 1 is characterized in that: The step of inputting the radial artery pulse wave signal into the trained CNN-BiLSTM model to predict the blood pressure signal further includes: Build a training dataset; Build a CNN-BiLSTM model; The CNN-BiLSTM model is trained and tested based on the training data set to obtain a trained CNN-BiLSTM model.

4. The radar blood pressure non-contact detection method based on the CNN-BiLSTM structure according to claim 3 is characterized in that: The process of constructing the training data set is as follows: The training dataset is constructed based on the fixed-length sliding window sample segmentation method.

5. The radar blood pressure non-contact detection method based on the CNN-BiLSTM structure according to claim 3 is characterized in that: The CNN-BiLSTM model includes a CNN network, a bidirectional long short-term memory network, a fully connected layer and an output layer, wherein the CNN network, the bidirectional long short-term memory network, the fully connected layer and the output layer are connected in sequence.

6. The radar blood pressure non-contact detection method based on the CNN-BiLSTM structure according to claim 3 is characterized in that: The process of training and testing the CNN-BiLSTM model based on the training data set to obtain the trained CNN-BiLSTM model is as follows: The CNN-BiLSTM model is trained and tested using Adam based on batch normalization and dropout technology to obtain a trained CNN-BiLSTM model.

7. A radar blood pressure non-contact detection system based on CNN-BiLSTM structure, characterized in that: include: An acquisition module, used for acquiring radial artery pulse wave signals; The prediction module is used to input the radial artery pulse wave signal into the trained CNN-BiLSTM model to predict the blood pressure signal and complete the radar blood pressure non-contact detection based on the CNN-BiLSTM structure.

8. The radar blood pressure non-contact detection system based on the CNN-BiLSTM structure according to claim 7 is characterized in that: Also includes: The first building module is used to build a training data set; The second building block is used to build the CNN-BiLSTM model; The training module is used to train and test the CNN-BiLSTM model based on the training data set to obtain a trained CNN-BiLSTM model.

9. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the computer program, the steps of the radar blood pressure non-contact detection method based on the CNN-BiLSTM structure as described in any one of claims 1 to 6 are implemented.

10. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the steps of the radar blood pressure non-contact detection method based on the CNN-BiLSTM structure as described in any one of claims 1 to 6 are implemented.

Images