Method and system for rapid vital sign analysis based on millimeter wave radar time series data

Abstract

The application provides a fast vital sign analysis method and system based on millimeter wave radar timing data, which comprises the following steps: obtaining a two-dimensional image corresponding to target millimeter wave radar data, the two-dimensional image comprising a distance dimension and a Doppler dimension, the target millimeter wave radar data being obtained by preprocessing noise data in millimeter wave radar data, the preprocessing comprising static clutter filtering, average maximum energy distance gate positioning and signal phase unwrapping; based on the two-dimensional image, obtaining a data cube corresponding to the target millimeter wave radar data, the data cube comprising a time dimension, a distance dimension and a Doppler dimension; projecting the data cube onto three orthogonal two-dimensional planes to obtain image data of the data cube on each two-dimensional plane, the three orthogonal two-dimensional planes being a time-distance plane, a time-Doppler plane and a distance-Doppler plane; and analyzing the image data to obtain a predicted value of vital sign data corresponding to the image data.

Description

Technical Field

[0001] This invention relates to the field of data processing technology, and in particular to a method and system for rapid vital sign analysis based on millimeter-wave radar time-series data. Background Technology

[0002] Millimeter-wave radar has been widely used in various fields, including autonomous vehicles, drone navigation, meteorological observation, and security monitoring. However, the high dimensionality and complexity of millimeter-wave radar data make its processing and analysis difficult.

[0003] Traditional millimeter-wave radar data processing and analysis methods, such as spectrum analysis, time spectrum analysis, and Doppler analysis, can extract information from millimeter-wave radar data to some extent, but they can only process one or two dimensions of the data, making it difficult to comprehensively present and utilize all the information in the data. Furthermore, these methods often require manual selection of specific parameters and thresholds, lacking flexibility and adaptability.

[0004] Therefore, there is an urgent need for a new method that can comprehensively and efficiently process and analyze millimeter-wave radar data, so as to make full use of this data and improve the performance of radar systems. Summary of the Invention

[0005] The present invention provides a rapid vital sign analysis method and system based on millimeter-wave radar time-series data, which is used to solve the problem of difficulty in comprehensively and efficiently processing and analyzing millimeter-wave radar data in the prior art.

[0006] This invention provides a rapid vital sign analysis method based on millimeter-wave radar time-series data, comprising:

[0007] A two-dimensional image corresponding to the target millimeter-wave radar data is obtained. The dimensions of the two-dimensional image include the range dimension and the Doppler dimension. The target millimeter-wave radar data is obtained by preprocessing the noise data in the millimeter-wave radar data. The preprocessing includes: static clutter filtering, average maximum energy range gate localization, and signal phase dewinding.

[0008] Based on the two-dimensional image, a data cube corresponding to the target millimeter-wave radar data is obtained. The dimensions of the data cube include time dimension, range dimension and Doppler dimension.

[0009] The data cube is projected onto three orthogonal two-dimensional planes to obtain image data of the data cube on each two-dimensional plane. The three orthogonal two-dimensional planes are the time-distance plane, the time-Doppler plane, and the distance-Doppler plane.

[0010] The image data is analyzed to obtain the predicted values ​​of the vital signs data corresponding to the image data.

[0011] In one embodiment, acquiring the two-dimensional image corresponding to the target millimeter-wave radar data includes:

[0012] Perform a one-dimensional fast Fourier transform on the target millimeter-wave radar data to obtain the range dimension data corresponding to the target millimeter-wave radar data;

[0013] The distance dimension data is subjected to a Fast Fourier Transform in the linear frequency modulation dimension to obtain the two-dimensional image.

[0014] In one embodiment, obtaining the data cube corresponding to the target millimeter-wave radar data based on the two-dimensional image includes:

[0015] Perform a two-dimensional Fourier transform on the two-dimensional image to obtain the image after the two-dimensional Fourier transform;

[0016] The images obtained by superimposing multiple frames of the two-dimensional Fourier transform on the time axis are used to obtain a three-dimensional image. The dimensions of the three-dimensional image include distance dimension, Doppler dimension and time dimension. The three-dimensional image is used to indicate the data cube.

[0017] In one embodiment, the step of obtaining a data cube corresponding to the target millimeter-wave radar data based on the two-dimensional image, and the step of projecting the data cube onto three orthogonal two-dimensional planes to obtain image data of the data cube on each two-dimensional plane, includes:

[0018] The projection onto the time-distance plane is obtained by summing or averaging the data corresponding to the Doppler dimension of the data cube.

[0019] The time dimension data and the distance dimension data corresponding to the data cube are summed respectively to obtain the projection on the distance Doppler plane and the projection on the time Doppler plane;

[0020] The projections on the time-distance plane, the distance-Doppler plane, and the time-Doppler plane are converted into image data of the data cube on each two-dimensional plane.

[0021] In one embodiment, analyzing the image data to obtain predicted values ​​of vital sign data corresponding to the image data includes:

[0022] The image data is input into the target Transformer model, and the image data is analyzed to obtain the predicted value of the vital signs data. The target Transformer model is obtained by training the Transformer model based on training samples. The training samples are obtained based on historical image data corresponding to historical millimeter-wave radar data. The historical millimeter-wave radar data is obtained by processing the noise data in the millimeter-wave radar data within the historical time period. The historical image data is the image data corresponding to the data cube corresponding to the historical millimeter-wave radar data.

[0023] This invention also provides a rapid vital sign analysis system based on millimeter-wave radar time-series data, comprising:

[0024] The first acquisition module is used to acquire a two-dimensional image corresponding to the target millimeter-wave radar data. The dimensions of the two-dimensional image include a range dimension and a Doppler dimension. The target millimeter-wave radar data is obtained by preprocessing the noise data in the millimeter-wave radar data. The preprocessing includes: static clutter filtering, average maximum energy range gate localization, and signal phase dewinding.

[0025] The second acquisition module is used to acquire a data cube corresponding to the target millimeter-wave radar data based on the two-dimensional image. The dimensions of the data cube include time dimension, distance dimension and Doppler dimension.

[0026] The processing module is used to project the data cube onto three orthogonal two-dimensional planes to obtain image data of the data cube on each two-dimensional plane. The three orthogonal two-dimensional planes are the time-distance plane, the time-Doppler plane, and the distance-Doppler plane.

[0027] The analysis module is used to analyze the image data and obtain the predicted values ​​of the vital signs data corresponding to the image data.

[0028] In some embodiments, the first acquisition module is specifically used for:

[0029] Perform a one-dimensional fast Fourier transform on the target millimeter-wave radar data to obtain the range dimension data corresponding to the target millimeter-wave radar data;

[0030] The distance dimension data is subjected to a Fast Fourier Transform in the linear frequency modulation dimension to obtain the two-dimensional image.

[0031] The present invention also provides an electronic device, including a processor and a memory storing a computer program, wherein the processor executes the program to implement the rapid vital sign analysis method based on millimeter-wave radar time-series data as described above.

[0032] The present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the rapid vital sign analysis method based on millimeter-wave radar time-series data as described above.

[0033] The present invention also provides a computer program product, including a computer program that, when executed by a processor, implements the rapid vital sign analysis method based on millimeter-wave radar time-series data as described above.

[0034] The present invention provides a rapid vital sign analysis method and system based on millimeter-wave radar time-series data. By converting the two-dimensional image corresponding to the target millimeter-wave radar data into a multi-dimensional data cube, the radar signal is analyzed from multiple directions, reducing noise and interference and improving the recognition accuracy of weak or mixed signals. By projecting the data cube onto three orthogonal two-dimensional planes in multiple dimensions, complex multi-dimensional data can be represented more intuitively, which helps to improve the intuitiveness and interpretability of data analysis. As a result, the predicted value of vital sign data corresponding to the image data can be obtained, improving the analysis accuracy and reliability of millimeter-wave radar data, realizing comprehensive and efficient processing and analysis of millimeter-wave radar data, and accurately predicting vital sign data. Attached Figure Description

[0035] To more clearly illustrate the technical solutions in this invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0036] Figure 1 This is one of the flowcharts of the rapid vital sign analysis method based on millimeter-wave radar time series data provided by the present invention;

[0037] Figure 2 This is the second flowchart of the rapid vital sign analysis method based on millimeter-wave radar time series data provided by the present invention;

[0038] Figure 3 This is a schematic diagram illustrating the analysis of target millimeter-wave radar data based on the target Transformer model provided by the present invention;

[0039] Figure 4 This is a schematic diagram of the structure of the rapid vital signs analysis system based on millimeter-wave radar time-series data provided by the present invention;

[0040] Figure 5 This is a schematic diagram of the physical structure of the electronic device provided by the present invention. Detailed Implementation

[0041] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this invention. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without creative effort are within the scope of protection of this invention.

[0042] Figure 1 This is one of the flowcharts illustrating the rapid vital sign analysis method based on millimeter-wave radar time-series data provided by the present invention, such as... Figure 1 As shown, the method includes:

[0043] Step 110: Obtain a two-dimensional image corresponding to the target millimeter-wave radar data. The dimensions of the two-dimensional image include the range dimension and the Doppler dimension. The target millimeter-wave radar data is obtained by preprocessing the noise data in the millimeter-wave radar data. The preprocessing includes: static clutter filtering, average maximum energy range gate localization, and signal phase unwrapping.

[0044] Step 120: Based on the two-dimensional image, obtain the data cube corresponding to the target millimeter-wave radar data. The dimensions of the data cube include time dimension, range dimension and Doppler dimension.

[0045] Step 130: Project the data cube onto three orthogonal two-dimensional planes to obtain the image data of the data cube on each two-dimensional plane. The three orthogonal two-dimensional planes are the time-distance plane, the time-Doppler plane, and the distance-Doppler plane.

[0046] Step 140: Analyze the image data to obtain the predicted values ​​of the vital signs data corresponding to the image data.

[0047] It should be noted that the above method can be implemented by computer equipment.

[0048] Optionally, the rapid vital sign analysis method based on millimeter-wave radar time-series data provided by the present invention can be applied to the analysis of various millimeter-wave radar data, such as frequency modulated continuous wave (FMCW) millimeter-wave radar data, multiple-input multiple-output (MIMO) radar data, phased array radar data, etc.

[0049] This invention converts frequency millimeter-wave radar data, multiple-input multiple-output radar data, and phase array radar data into time-range-Doppler data cubes, and then uses a three-dimensional orthogonal projection method for analysis and visualization.

[0050] Specifically, target millimeter-wave radar data can be obtained by preprocessing the noise data of the aforementioned millimeter-wave radar data. Preprocessing includes, but is not limited to, static clutter filtering, average maximum energy range gating, signal phase dewinding, interpolation, denoising, and normalization. Preprocessing the raw millimeter-wave radar data can improve its quality and usability.

[0051] The following sections will explain static clutter filtering, average maximum energy distance gate localization, and signal phase unwinding.

[0052] In practice, static clutter filtering is performed on the raw millimeter-wave radar data.

[0053] The echo information collected by millimeter-wave radar includes target information and surrounding static environment information. In order to effectively extract the target signal, the static background signal reflected by the surrounding environment needs to be filtered out.

[0054] Static clutter filtering consists of two steps. The first step involves acquiring F frames of static background signal and averaging them in the absence of personnel in the radar detection area. This average is then used as the reference static clutter signal R. c (m).

[0055]

[0056] Where n is the fast-time sampling index, m is the slow-time sampling index, and R t-n This is for radar echo data when there are no targets in the detection area.

[0057] The second step is the target signal detection stage, which involves collecting each frame of raw radar signal R. t Subtract the reference static clutter signal R obtained in the previous step from all of them. c This leads to the acquisition of a signal R containing the target. h .

[0058] R h (n,m)=R t (n,m)-R c (m);

[0059] In practice, the filtered millimeter-wave radar data is subjected to average maximum energy range-gated positioning to determine the effective target signal.

[0060] During the detection process, the target's position relative to the radar cannot be completely stationary. Therefore, it is necessary to determine the target's distance relative to the radar, i.e., to locate the position of the effective vibration information in the fast time dimension. Frames with large target motion amplitude and positional shift will have a certain impact on the detection, so they are discarded as invalid frames. For frame data segments with slight target movement and relative stillness, the slow-time signal corresponding to the average maximum energy range gate w of the preceding and following h frames is used as the effective target signal, i.e.:

[0061] w = arg n maxE(n,m);

[0062] Where w is the value of n corresponding to the maximum value of the energy E(n,m) corresponding to the average maximum energy distance gate. The energy corresponding to the average maximum energy distance gate, that is, the energy E(n,m) corresponding to each sampling point in the fast time dimension of each frame, is calculated by the following formula:

[0063]

[0064] Since radar signals are acquired by inversely solving for phase changes, and the phase information requires the arctangent function for extraction, the arctangent function is often used. When solving for the arctangent function, angles in the first and second quadrants are assumed to be from 0 to π, while angles in the third and fourth quadrants are considered to be from 0 to -π. Therefore, the actual result will show a phase jump with an amplitude of 2π, a phenomenon called phase winding. During signal extraction, phase winding affects the extraction quality; therefore, the extracted phase needs to be unwound to prevent a phase jump at π, thus obtaining the true phase change.

[0065] Whether to unwind is determined by the difference between two adjacent moments on the slow-time sampling axis. Assume the vibration phase time sequence before unwinding is φ. b,i (n,m), the vibration phase time sequence after unwinding is φ C (n,m). Define the unwinding threshold as T=π. For the same set of vibration sampling sequences, if |φ b,i (n,m)-φ b,i If (n,m-1)∣>T, then unwind according to the following formula.

[0066] φ C (n,m)=φ b,i (n,m)+2kπ

[0067]

[0068] After the millimeter-wave radar data is preprocessed, the target millimeter-wave radar data is obtained and then converted into a Time-Range-Doppler (TRD) data cube.

[0069] The TRD data cube is a three-dimensional data structure where each dimension represents an important physical quantity. Time represents the transmission and reception time of the radar signal, distance represents the distance from the target object (i.e., the object being measured) to the radar, and Doppler represents the velocity of the target object relative to the radar. The TRD data cube contains all the information from the target millimeter-wave radar data and provides rich resources for subsequent data processing and analysis.

[0070] Based on this, a three-dimensional fast Fourier transform can be used to obtain the TRD data cube corresponding to the target millimeter-wave radar data.

[0071] It should be noted that 3D orthographic projection is a geometric transformation used to project data or objects in 3D space onto a 2D plane for easier observation and analysis. A key characteristic of this projection is that it preserves many important properties of the original data, such as length, angle, and shape.

[0072] In this invention, three-dimensional orthogonal projection is used to project the TRD data cube onto three orthogonal two-dimensional planes: the time-distance plane, the distance-Doppler plane, and the time-Doppler plane. The resulting projections on the two-dimensional planes are then converted into corresponding image data.

[0073] After obtaining the image data, a trained model can be used to analyze it.

[0074] For example, the target millimeter-wave radar data can be obtained by processing the noise data in the FMCW millimeter-wave radar data, and the TRD data cube corresponding to the target millimeter-wave radar data can be obtained based on the three-dimensional fast Fourier transform. The image data corresponding to the TRD data cube can be obtained based on the three-dimensional orthogonal projection. The image data is then input into the trained Transformer model (i.e., the target Transformer model) and analyzed to obtain the predicted values ​​of the vital signs data corresponding to the image data. The vital signs data can be the heart rate and respiratory rate of the measured object.

[0075] The present invention provides a rapid vital sign analysis method based on millimeter-wave radar time-series data. By converting the two-dimensional image corresponding to the target millimeter-wave radar data into a multi-dimensional data cube, the radar signal is analyzed from multiple directions, reducing noise and interference and improving the recognition accuracy of weak or mixed signals. By projecting the data cube onto three orthogonal two-dimensional planes in multiple dimensions, complex multi-dimensional data can be represented more intuitively, which helps to improve the intuitiveness and interpretability of data analysis. As a result, the predicted value of vital sign data corresponding to the image data can be obtained, improving the analysis accuracy and reliability of millimeter-wave radar data. This enables comprehensive and efficient processing and analysis of millimeter-wave radar data, and can accurately predict vital sign data.

[0076] In one embodiment, acquiring a two-dimensional image corresponding to the target millimeter-wave radar data includes:

[0077] Perform a one-dimensional fast Fourier transform on the target millimeter-wave radar data to obtain the range dimension data corresponding to the target millimeter-wave radar data;

[0078] A two-dimensional image is obtained by performing a Fast Fourier Transform on the distance dimension data in the linear frequency modulation dimension.

[0079] In practice, Fourier transform is used to process radar signals in order to separate different signals. The spectrum generated by the Fourier transform will produce different peaks for different signals, and each peak represents an object at a specific distance.

[0080] First, perform a one-dimensional fast fourier transform (1D FFT) on each column of the target millimeter-wave radar data, which is a fast fourier transform of the range dimension, to obtain the radar's range dimension data.

[0081] Performing a 1D FFT on the distance dimension data in the linear frequency modulation dimension (chirp dimension), i.e., the fast Fourier transform in the Doppler dimension, yields a two-dimensional image, i.e., a distance-Doppler image.

[0082] In one embodiment, obtaining the data cube corresponding to the target millimeter-wave radar data based on a two-dimensional image includes:

[0083] Perform a two-dimensional Fourier transform on the two-dimensional image to obtain the image after the two-dimensional Fourier transform;

[0084] By superimposing multiple frames of two-dimensional Fourier transform images on the time axis, a three-dimensional image is obtained. The dimensions of the three-dimensional image include distance dimension, Doppler dimension and time dimension. The three-dimensional image is used to indicate the data cube.

[0085] In practice, each frame of the two-dimensional image is processed by 2D FFT, and the processed data is stored frame by frame to obtain the two-dimensional Fourier transform image. Then, the two-dimensional Fourier transform images of multiple frames are superimposed in the time dimension to obtain a three-dimensional image, which is the distance-Doppler-time image.

[0086] Since the radar acquisition time for each frame is fixed, the number of frames corresponds to the corresponding time. The resulting data cube is a time-range-Doppler data cube.

[0087] In addition, other data processing steps are usually required to improve the quality of radar data, such as window function processing and clutter suppression.

[0088] The rapid vital sign analysis method based on millimeter-wave radar time-series data provided by this invention utilizes Fast Fourier Transform to convert target millimeter-wave radar data into a time-range-Doppler data cube, providing a more comprehensive and richer data structure that allows for analysis of radar signals from multiple directions. This helps reduce noise and interference, and improves the accuracy of identifying weak or mixed signals. It enables more accurate and reliable monitoring of the heart rate and respiratory rate of the measured object.

[0089] Furthermore, in one embodiment, the data cube is projected onto three orthogonal two-dimensional planes to obtain image data of the data cube on each two-dimensional plane, including:

[0090] Summing or averaging the data in the Doppler dimension corresponding to the data cube yields the projection onto the time-distance plane.

[0091] Summing the data in the time dimension and the data in the distance dimension corresponding to the data cube respectively yields the projections on the distance Doppler plane and the time Doppler plane;

[0092] The projections on the time-distance plane, the distance-Doppler plane, and the time-Doppler plane are converted into image data of the data cube on each two-dimensional plane.

[0093] Optionally, three-dimensional orthogonal projection is used to project the TRD data cube onto three orthogonal two-dimensional planes: the time-distance plane, the distance-Doppler plane, and the time-Doppler plane.

[0094] Specifically, time-distance projection: This projection process is based on three-dimensional orthogonal projection, summing or averaging the data in the Doppler dimensions of the TRD data cube. The result is the projection onto the time-distance plane. The value of each point on the time-distance plane is equal to the sum or average of all corresponding Doppler values ​​in the TRD data cube. This projection describes the changes of the target object in the time and distance dimensions.

[0095] The distance-Doppler projection is based on summing the time-dimension data of the TRD data cube using a three-dimensional orthogonal projection, resulting in a projection onto the distance-Doppler plane. The value of each point on the distance-Doppler plane is equal to the sum of all corresponding time values ​​in the TRD data cube.

[0096] The time-Doppler projection is a projection onto the time-Doppler plane obtained by summing the distance dimensions of the TRD data cube based on a three-dimensional orthogonal projection. The value of each point on the time-Doppler plane is equal to the sum of all corresponding distance values ​​in the TRD data cube.

[0097] These three projections project a three-dimensional data cube onto a two-dimensional plane, greatly simplifying the complexity of data analysis and making the results more intuitive and easier to understand. It's important to note that while this projection method reduces the dimensionality of the data to some extent, it may also lead to the loss of some information. Therefore, choosing appropriate projection directions and methods to preserve as much information as possible from the original data is a crucial issue to consider in this process.

[0098] The projections on the three two-dimensional planes (the projection on the time-range plane, the projection on the range-Doppler plane) are converted into corresponding image data, resulting in three image data sets. These images are convenient for human observation and analysis, as well as for automated processing and analysis by computers. These three image data sets represent the time-range, range-Doppler, and time-Doppler relationships, respectively, reflecting the characteristics of the target millimeter-wave radar data from different perspectives.

[0099] Specifically, converting the obtained projection data on the two-dimensional plane into image data can be achieved through several image processing techniques. Below is a common method:

[0100] Normalization: First, the two-dimensional projection data is normalized. Normalization maps the data range to between 0 and 255 for subsequent image display.

[0101] Converting to grayscale: This involves mapping the normalized data to pixel intensity values ​​in a grayscale image. Depending on the data's numerical range and the chosen grayscale level, smaller values ​​can be mapped to darker grayscale values, and larger values ​​to brighter grayscale values.

[0102] Image generation: Based on the normalized and grayscale data, pixels are filled into their corresponding positions according to their location in a two-dimensional projection to generate an image. This step can be implemented using image processing libraries or image processing functions in a programming language.

[0103] Visualization: Finally, the generated image is visualized. This can be done using image viewers, image processing software, or image display functions in programming languages.

[0104] It is important to note that the generated image may contain some distortion or noise, depending on the quality of the original 2D projection data and the chosen processing method. Therefore, in practical applications, some image enhancement or denoising steps may be necessary to optimize the image quality and sharpness.

[0105] The rapid vital sign analysis method based on millimeter-wave radar time-series data provided by this invention utilizes three-dimensional orthogonal projection analysis of the time-range-Doppler data cube (TRD). This allows for comprehensive and multi-angle analysis of target millimeter-wave radar data, capturing crucial information that might be overlooked from a single viewpoint. By projecting the three-dimensional TRD data cube onto a two-dimensional plane, complex multi-dimensional data can be represented more intuitively, improving the intuitiveness and interpretability of data analysis. Furthermore, orthogonal projection ensures independence and distortion-free operation between different dimensions, facilitating independent data analysis in different directions and enhancing analysis accuracy and reliability.

[0106] Furthermore, in one embodiment, the image data is analyzed to obtain predicted values ​​of vital sign data corresponding to the image data, including:

[0107] Image data is input into the target Transformer model, and the image data is analyzed to obtain the predicted values ​​of vital signs data. The target Transformer model is obtained by training the Transformer model based on training samples. The training samples are obtained based on historical image data corresponding to historical millimeter-wave radar data. The historical millimeter-wave radar data is obtained by processing the noise data in the millimeter-wave radar data within the historical time period. The historical image data is the image data corresponding to the data cube of the historical millimeter-wave radar data.

[0108] Optionally, the image data can be analyzed in conjunction with the Transformer model. The obtained image data can be input into the target Transformer model to analyze the image data corresponding to the target millimeter-wave radar data in real time.

[0109] The target Transformer model is obtained by training the Transformer model based on the training samples.

[0110] The training samples are obtained based on historical image data corresponding to historical millimeter-wave radar data. This historical millimeter-wave radar data is obtained by processing the noise data in the millimeter-wave radar data within the historical time period. The historical image data is the image data corresponding to the time-range-Doppler data cube of the historical millimeter-wave radar data obtained by three-dimensional orthogonal projection.

[0111] It should be noted that the specific process of obtaining the image data corresponding to the historical millimeter-wave radar data is the same as the specific process of obtaining the image data of the millimeter-wave radar data, and will not be described again in this invention.

[0112] The following is the training process for the Transformer model:

[0113] 1. Input three historical image data (time-range image, time-Doppler image, and range-Doppler image, obtained by projecting the TRD data cubes corresponding to the historical millimeter-wave radar data onto the corresponding two-dimensional planes based on three-dimensional orthogonal projection) into the Transformer model.

[0114] 2. Train the Transformer model;

[0115] a) Feature extraction:

[0116] A series of convolutional layers are used to extract features from each image patch (obtained after segmenting three historical image data). These features will be used as input to the encoder part of the Transformer.

[0117] b) Location encoding:

[0118] Since the Transformer model itself does not contain any order information, positional encoding must be added to encode the positional information of image patches in the image.

[0119] This can be achieved by adding the location encoding to the feature representation of the image patch.

[0120] c) Transformer encoder:

[0121] A self-attention mechanism is used to capture dependencies between image patches. Similarity scores are calculated between each image patch and all other image patches, and these scores are then used to weight and combine the features of the image patches. Multi-head attention allows the model to focus on different aspects of the image simultaneously. Each "head" learns a different attention distribution. After each encoder layer, there is a feedforward neural network responsible for performing pointwise nonlinear transformations. Finally, residual connections and layer normalization are used to stabilize the training process and accelerate convergence.

[0122] The trained Transformer model, as the target Transformer model, can analyze new target millimeter-wave radar data in real time.

[0123] Depending on the specific application scenario, the Transformer model can be further optimized:

[0124] a) Adjust the model architecture: You can add or reduce the number of layers, change the number of hidden units, etc., to adapt to the needs of specific tasks.

[0125] b) Parallel processing: Parallel computing is achieved by using multi-head attention mechanisms and graphics processing units (GPUs) to accelerate processing speed.

[0126] c) Quantization and compression: Reduce the model size through model quantization and compression techniques to make it more suitable for running on resource-constrained devices.

[0127] The rapid vital sign analysis method based on millimeter-wave radar time-series data provided by this invention differs from traditional radar preprocessing methods, which may involve multiple steps and models. This invention uses a Transformer model to provide an end-to-end solution. This means that training can begin directly from the raw data, ultimately predicting heart rate and respiratory rate directly, reducing the complexity and potential errors of intermediate steps. Furthermore, the Transformer model can process the entire data sequence simultaneously, rather than stepwise, which greatly improves computational efficiency. Simultaneously, its modular structure allows for customization and optimization to adapt to specific millimeter-wave radar configurations or specific monitoring needs.

[0128] Furthermore, in one embodiment, the method of acquiring target millimeter-wave radar data may further include:

[0129] The noise data in the millimeter-wave radar data is filtered and denoised to obtain the first millimeter-wave radar data.

[0130] The first millimeter-wave radar data is normalized to obtain the target millimeter-wave radar data.

[0131] Optionally, the corresponding target millimeter-wave radar data can be obtained by preprocessing the millimeter-wave radar data.

[0132] Specifically, data preprocessing is a preparatory step that processes raw millimeter-wave radar data to improve its quality and usability. Specific operations include filtering, interpolation, denoising, and normalization.

[0133] Filtering is a common data preprocessing method that helps remove noise and interference from data, retaining only the desired signal. For example, a low-pass filter can be used to remove high-frequency noise from millimeter-wave radar data to obtain the first millimeter-wave radar data.

[0134] Denoising is used to eliminate noise in millimeter-wave radar data, which can typically be achieved through methods such as smoothing and median filtering.

[0135] Normalization transforms all data values ​​to a uniform range, such as 0 to 1 or -1 to 1, which facilitates subsequent data processing and analysis. For example, by normalizing the first millimeter-wave radar data, the corresponding target millimeter-wave radar data can be obtained.

[0136] Figure 2 This is the second flowchart of the rapid vital sign analysis method based on millimeter-wave radar time-series data provided by the present invention, as shown below. Figure 2 As shown, it includes:

[0137] S1, Data Preprocessing; S2, Data Transformation; S3, 3D Orthogonal Projection; S4, Image Transformation; and S5, using a target Transformer model to analyze image data to predict heart rate and respiratory rate. The data preprocessing stage includes filtering, denoising, and normalization to improve data quality and usability. The data transformation step converts the preprocessed radar data into a Time-Range-Doppler (TRD) data cube. 3D orthogonal projection projects this data cube onto three orthogonal two-dimensional planes for easier observation and analysis. Image transformation converts the two-dimensional projection into an image, facilitating human observation and computer processing. The final step combines the Transformer model, achieving real-time analysis and prediction through training and optimization.

[0138] Figure 3 This is a schematic diagram illustrating the analysis of target millimeter-wave radar data based on the target Transformer model provided by the present invention, as shown below. Figure 3 As shown, it includes:

[0139] Target millimeter-wave radar data is converted into a TRD data cube using 3D fast FFT;

[0140] The data cube is projected onto three orthogonal two-dimensional planes using three-dimensional orthogonal projection: the time-distance plane, the distance-Doppler plane, and the time-Doppler plane, to obtain three two-dimensional image data.

[0141] Three two-dimensional image data are input into the target Transformer model for analysis to obtain the heart rate and respiratory rate of the measured object.

[0142] The rapid vital sign analysis method based on millimeter-wave radar time-series data provided by this invention can comprehensively utilize FMCW millimeter-wave radar data to extract richer information. Furthermore, by projecting the TRD data cube onto three orthogonal planes, the radar data can be understood and analyzed more intuitively.

[0143] Furthermore, this invention can process large amounts of millimeter-wave radar data with high efficiency; it can be applied to various types of FMCW millimeter-wave radar, exhibiting strong versatility; its operation is simple, easy to implement, and conducive to promotion and application.

[0144] This invention provides a new and effective system and method for processing and analyzing FMCW millimeter-wave radar data, which has important practical value for improving the performance of radar systems.

[0145] The rapid vital sign analysis method based on millimeter-wave radar time-series data provided by this invention preprocesses the millimeter-wave radar data before analysis to remove noise data and normalizes the data to a uniform range, thereby improving data quality and usability.

[0146] The rapid vital signs analysis method based on millimeter-wave radar time-series data provided by the present invention is described below. The rapid vital signs analysis method based on millimeter-wave radar time-series data described below can be referred to in correspondence with the rapid vital signs analysis method based on millimeter-wave radar time-series data described above.

[0147] Figure 4 This is a schematic diagram of the structure of the rapid vital sign analysis system based on millimeter-wave radar time-series data provided by the present invention, as shown below. Figure 4 As shown, it includes:

[0148] The first acquisition module 410 is used to acquire a two-dimensional image corresponding to the target millimeter-wave radar data. The dimensions of the two-dimensional image include a range dimension and a Doppler dimension. The target millimeter-wave radar data is obtained by preprocessing the noise data in the millimeter-wave radar data. The preprocessing includes: static clutter filtering, average maximum energy range gate localization, and signal phase dewinding.

[0149] The second acquisition module 411 is used to acquire a data cube corresponding to the target millimeter-wave radar data based on the two-dimensional image. The dimensions of the data cube include time dimension, distance dimension and Doppler dimension.

[0150] Processing module 412 is used to project the data cube onto three orthogonal two-dimensional planes to obtain image data of the data cube on each two-dimensional plane. The three orthogonal two-dimensional planes are a time-distance plane, a time-Doppler plane, and a distance-Doppler plane.

[0151] The analysis module 413 is used to analyze the image data to obtain the predicted value of the vital signs data corresponding to the image data.

[0152] In one embodiment, the first acquisition module 410 is specifically used for:

[0153] Perform a one-dimensional fast Fourier transform on the target millimeter-wave radar data to obtain the range dimension data corresponding to the target millimeter-wave radar data;

[0154] The distance dimension data is subjected to a Fast Fourier Transform in the linear frequency modulation dimension to obtain the two-dimensional image.

[0155] In one embodiment, the second acquisition module 411 is specifically used for:

[0156] Perform a two-dimensional Fourier transform on the two-dimensional image to obtain the image after the two-dimensional Fourier transform;

[0157] The images obtained by superimposing multiple frames of the two-dimensional Fourier transform on the time axis are used to obtain a three-dimensional image. The dimensions of the three-dimensional image include distance dimension, Doppler dimension and time dimension. The three-dimensional image is used to indicate the data cube.

[0158] In one embodiment, the processing module 412 is specifically used for:

[0159] The projection onto the time-distance plane is obtained by summing or averaging the data corresponding to the Doppler dimension of the data cube.

[0160] The time dimension data and the distance dimension data corresponding to the data cube are summed respectively to obtain the projection on the distance Doppler plane and the projection on the time Doppler plane;

[0161] The projections on the time-distance plane, the distance-Doppler plane, and the time-Doppler plane are converted into image data of the data cube on each two-dimensional plane.

[0162] In one embodiment, the analysis module 413 is specifically used for:

[0163] The image data is input into the target Transformer model, and the image data is analyzed to obtain the predicted value of the vital signs data. The target Transformer model is obtained by training the Transformer model based on training samples. The training samples are obtained based on historical image data corresponding to historical millimeter-wave radar data. The historical millimeter-wave radar data is obtained by processing the noise data in the millimeter-wave radar data within the historical time period. The historical image data is the image data corresponding to the data cube corresponding to the historical millimeter-wave radar data.

[0164] The rapid vital sign analysis system based on millimeter-wave radar time-series data provided by this invention transforms the two-dimensional image corresponding to the target millimeter-wave radar data into a multi-dimensional data cube, analyzes radar signals from multiple directions, reduces noise and interference, and improves the recognition accuracy of weak or mixed signals. By projecting the data cube onto three orthogonal two-dimensional planes in multiple dimensions, complex multi-dimensional data can be represented more intuitively, which helps to improve the intuitiveness and interpretability of data analysis. As a result, the predicted value of vital sign data corresponding to the image data can be obtained, improving the accuracy and reliability of millimeter-wave radar data analysis, realizing comprehensive and efficient processing and analysis of millimeter-wave radar data, and accurately predicting vital sign data.

[0165] Figure 5 This is a schematic diagram of the physical structure of an electronic device provided by the present invention, such as... Figure 5 As shown, the electronic device may include a processor 510, a communication interface 511, a memory 512, and a communication bus 513. The processor 510, communication interface 511, and memory 512 communicate with each other via the communication bus 513. The processor 510 can call logical instructions from the memory 512 to execute the following methods:

[0166] A two-dimensional image corresponding to the target millimeter-wave radar data is obtained. The dimensions of the two-dimensional image include the range dimension and the Doppler dimension. The target millimeter-wave radar data is obtained by preprocessing the noise data in the millimeter-wave radar data. The preprocessing includes: static clutter filtering, average maximum energy range gate localization, and signal phase dewinding.

[0167] Based on the two-dimensional image, a data cube corresponding to the target millimeter-wave radar data is obtained. The dimensions of the data cube include time dimension, range dimension and Doppler dimension.

[0168] The data cube is projected onto three orthogonal two-dimensional planes to obtain image data of the data cube on each two-dimensional plane. The three orthogonal two-dimensional planes are the time-distance plane, the time-Doppler plane, and the distance-Doppler plane.

[0169] The image data is analyzed to obtain the predicted values ​​of the vital signs data corresponding to the image data.

[0170] Furthermore, the logical instructions in the aforementioned memory can be implemented as software functional units and sold or used as independent products, and can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer power supply (which may be a personal computer, server, or network power supply, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0171] Furthermore, this invention discloses a computer program product, which includes a computer program stored on a non-transitory computer-readable storage medium. The computer program includes program instructions, and when these instructions are executed by a computer, the computer can execute the rapid vital sign analysis method based on millimeter-wave radar time-series data provided in the above-described method embodiments, for example including:

[0172] A two-dimensional image corresponding to the target millimeter-wave radar data is obtained. The dimensions of the two-dimensional image include the range dimension and the Doppler dimension. The target millimeter-wave radar data is obtained by preprocessing the noise data in the millimeter-wave radar data. The preprocessing includes: static clutter filtering, average maximum energy range gate localization, and signal phase dewinding.

[0173] Based on the two-dimensional image, a data cube corresponding to the target millimeter-wave radar data is obtained. The dimensions of the data cube include time dimension, range dimension and Doppler dimension.

[0174] The data cube is projected onto three orthogonal two-dimensional planes to obtain image data of the data cube on each two-dimensional plane. The three orthogonal two-dimensional planes are the time-distance plane, the time-Doppler plane, and the distance-Doppler plane.

[0175] The image data is analyzed to obtain the predicted values ​​of the vital signs data corresponding to the image data.

[0176] On the other hand, the present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the rapid vital sign analysis method based on millimeter-wave radar time-series data provided in the above embodiments, including, for example:

[0177] A two-dimensional image corresponding to the target millimeter-wave radar data is obtained. The dimensions of the two-dimensional image include the range dimension and the Doppler dimension. The target millimeter-wave radar data is obtained by preprocessing the noise data in the millimeter-wave radar data. The preprocessing includes: static clutter filtering, average maximum energy range gate localization, and signal phase dewinding.

[0178] Based on the two-dimensional image, a data cube corresponding to the target millimeter-wave radar data is obtained. The dimensions of the data cube include time dimension, range dimension and Doppler dimension.

[0179] The data cube is projected onto three orthogonal two-dimensional planes to obtain image data of the data cube on each two-dimensional plane. The three orthogonal two-dimensional planes are the time-distance plane, the time-Doppler plane, and the distance-Doppler plane.

[0180] The image data is analyzed to obtain the predicted values ​​of the vital signs data corresponding to the image data.

[0181] The system embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.

[0182] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., including several instructions to cause a computer power supply (which may be a personal computer, server, or network power supply, etc.) to execute the methods described in various embodiments or some parts of the embodiments.

[0183] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A rapid vital sign analysis method based on millimeter-wave radar time-series data, characterized in that, include: A two-dimensional image corresponding to the target millimeter-wave radar data is obtained. The dimensions of the two-dimensional image include the range dimension and the Doppler dimension. The target millimeter-wave radar data is obtained by preprocessing the noise data in the millimeter-wave radar data. The preprocessing includes: static clutter filtering, average maximum energy range gate localization, and signal phase dewinding. Based on the two-dimensional image, a data cube corresponding to the target millimeter-wave radar data is obtained. The dimensions of the data cube include time dimension, range dimension and Doppler dimension. The data cube is projected onto three orthogonal two-dimensional planes to obtain image data of the data cube on each two-dimensional plane. The three orthogonal two-dimensional planes are the time-distance plane, the time-Doppler plane, and the distance-Doppler plane. Analyzing the image data to obtain predicted values ​​of the corresponding vital signs data includes: The image data is input into the target Transformer model, and the image data is analyzed to obtain the predicted value of the vital signs data. The target Transformer model is obtained by training the Transformer model based on training samples. The training samples are obtained based on historical image data corresponding to historical millimeter-wave radar data. The historical millimeter-wave radar data is obtained by processing the noise data in the millimeter-wave radar data within the historical time period. The historical image data is the image data corresponding to the data cube corresponding to the historical millimeter-wave radar data.

2. The rapid vital sign analysis method based on millimeter-wave radar time-series data according to claim 1, characterized in that, The acquisition of the two-dimensional image corresponding to the target millimeter-wave radar data includes: Perform a one-dimensional fast Fourier transform on the target millimeter-wave radar data to obtain the range dimension data corresponding to the target millimeter-wave radar data; The distance dimension data is subjected to a Fast Fourier Transform in the linear frequency modulation dimension to obtain the two-dimensional image.

3. The rapid vital sign analysis method based on millimeter-wave radar time-series data according to claim 1, characterized in that, The step of obtaining the data cube corresponding to the target millimeter-wave radar data based on the two-dimensional image includes: Perform a two-dimensional Fourier transform on the two-dimensional image to obtain the image after the two-dimensional Fourier transform; The images obtained by superimposing multiple frames of the two-dimensional Fourier transform on the time axis are used to obtain a three-dimensional image. The dimensions of the three-dimensional image include distance dimension, Doppler dimension and time dimension. The three-dimensional image is used to indicate the data cube.

4. The rapid vital sign analysis method based on millimeter-wave radar time-series data according to claim 1, characterized in that, The step of projecting the data cube onto three orthogonal two-dimensional planes to obtain image data of the data cube on each two-dimensional plane includes: The projection onto the time-distance plane is obtained by summing or averaging the data corresponding to the Doppler dimension of the data cube. The time dimension data and the distance dimension data corresponding to the data cube are summed respectively to obtain the projection on the distance Doppler plane and the projection on the time Doppler plane; The projections on the time-distance plane, the distance-Doppler plane, and the time-Doppler plane are converted into image data of the data cube on each two-dimensional plane.

5. A rapid vital sign analysis system based on millimeter-wave radar time-series data, characterized in that, include: The first acquisition module is used to acquire a two-dimensional image corresponding to the target millimeter-wave radar data. The dimensions of the two-dimensional image include a range dimension and a Doppler dimension. The target millimeter-wave radar data is obtained by preprocessing the noise data in the millimeter-wave radar data. The preprocessing includes: static clutter filtering, average maximum energy range gate localization, and signal phase dewinding. The second acquisition module is used to acquire a data cube corresponding to the target millimeter-wave radar data based on the two-dimensional image. The dimensions of the data cube include time dimension, distance dimension and Doppler dimension. The processing module is used to project the data cube onto three orthogonal two-dimensional planes to obtain image data of the data cube on each two-dimensional plane. The three orthogonal two-dimensional planes are the time-distance plane, the time-Doppler plane, and the distance-Doppler plane. The analysis module is used to analyze the image data to obtain the predicted values ​​of the vital signs data corresponding to the image data; The analysis module is specifically used for: The image data is input into the target Transformer model, and the image data is analyzed to obtain the predicted value of the vital signs data. The target Transformer model is obtained by training the Transformer model based on training samples. The training samples are obtained based on historical image data corresponding to historical millimeter-wave radar data. The historical millimeter-wave radar data is obtained by processing the noise data in the millimeter-wave radar data within the historical time period. The historical image data is the image data corresponding to the data cube corresponding to the historical millimeter-wave radar data.

6. The rapid vital sign analysis system based on millimeter-wave radar time-series data according to claim 5, characterized in that, The first acquisition module is specifically used for: Perform a one-dimensional fast Fourier transform on the target millimeter-wave radar data to obtain the range dimension data corresponding to the target millimeter-wave radar data; The distance dimension data is subjected to a Fast Fourier Transform in the linear frequency modulation dimension to obtain the two-dimensional image.

7. An electronic device comprising a processor and a memory storing a computer program, characterized in that, When the processor executes the computer program, it implements the rapid vital sign analysis method based on millimeter-wave radar time-series data as described in any one of claims 1 to 4.

8. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that, When executed by a processor, the computer program implements the rapid vital signs analysis method based on millimeter-wave radar time-series data as described in any one of claims 1 to 4.

9. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by the processor, it implements the rapid vital sign analysis method based on millimeter-wave radar time-series data as described in any one of claims 1 to 4.

Images