Respiratory monitoring methods, media, devices and systems based on ultra-wideband radio frequency signals

Abstract

This invention discloses a respiratory monitoring method, medium, device, and system based on ultra-wideband radio frequency signals. It acquires signal data returned by IR-UWB radar; performs FFT on the signal in each interval and multiplies the spectra of three adjacent intervals, using the maximum coefficient within the corresponding frequency range as a metric to locate the interval in which the human is breathing; applies a similarity transformation method to the signal within the human breathing interval to eliminate background noise; compensates for phase shifts caused by the monitored person's position and orientation, and normalizes the signal amplitude to eliminate the influence of irrelevant factors such as the monitored person's position and orientation; further feeds the amplitude, I-component, and Q-component of the signal in two breathing intervals into a trained encoder-decoder neural network, outputting a respiratory waveform through the encoder-decoder neural network, thus achieving non-contact monitoring of human respiration, which can be widely applied in the medical field.

Description

Technical Field

[0001] This invention belongs to the field of respiratory monitoring technology, specifically relating to a respiratory monitoring method, medium, device, and system based on ultra-wideband radio frequency signals. Background Technology

[0002] Automated monitoring of patient responsiveness is a crucial aspect of medical procedures. Standard medical equipment, such as cardiac and respiratory monitors and electrocardiogram (ECG) devices, is commonly used in hospitals. While these devices are accurate, most are expensive and can be somewhat disruptive to patients. Furthermore, they may not be suitable for patients with burns or trauma. Considering these issues, some researchers have proposed using wireless methods to detect and extract vital signs. Wireless-based vital sign monitoring is non-contact, economical, and easy to deploy.

[0003] However, the propagation characteristics of wireless sensing make it susceptible to environmental factors and human posture when receiving signals, introducing unpredictable errors into vital sign monitoring. Previous research has mainly proposed two methods to address these issues: knowledge-driven methods and data-driven neural networks. The former constructs an appropriate propagation model to estimate and eliminate environmental factors. The latter uses neural networks to learn the intrinsic relationship between the feature space and the noise space. However, both have limitations: knowledge-driven models assume a relatively stable environment, which cannot be guaranteed in many hospital environments; data-driven methods require large amounts of data for training, and the trained models are limited to specific environments. Summary of the Invention

[0004] The technical problem to be solved by this invention is to address the shortcomings of the prior art by providing a respiratory monitoring method, medium, device and system based on ultra-wideband radio frequency signals. It uses commercial pulsed wireless ultra-wideband (IR-UWB) radar and combines knowledge-driven and data-driven methods to eliminate irrelevant environmental factors and restore the human respiratory waveform, thereby achieving non-contact monitoring of human respiration. This solves the technical problem that wireless signals cannot be used for vital sign monitoring in indoor scenarios.

[0005] The present invention adopts the following technical solution:

[0006] A respiratory monitoring method based on ultra-wideband radio frequency signals, comprising the following steps:

[0007] S1. Use pulse wireless ultra-wideband radar to send pulse signals to sense human breathing and the environment, obtain a signal matrix, and put each data of the signal matrix along the time axis into the complex plane to obtain the corresponding I / Q constellation diagram.

[0008] S2. Perform a fast Fourier transform on each distance interval of the signal matrix obtained in step S1 along the time direction, and then multiply the spectral features of three adjacent consecutive intervals. Based on the maximum value within the respiratory frequency range, locate the distance interval of the user's breathing in the signal matrix.

[0009] S3. The signal matrix within the distance interval where the user's breathing is located, obtained in step S2, is modeled as the sum of the background reflection offset signal and the breathing-inducing signal. Based on the I / Q constellation diagram obtained in step S1, the background reflection offset signal is removed by the adjacent distance interval similarity transformation method to obtain the breathing-inducing signal.

[0010] S4. Process the respiratory evoked signal obtained in step S3 to compensate for the phase shift caused by the position and direction of the monitored person, and normalize the amplitude of the respiratory evoked signal to obtain two normalized respiratory distance interval signals.

[0011] S5. Design and train an encoder-decoder neural network based on U-Net. Input the amplitude, real part I and imaginary part Q components of the two breathing distance interval signals after normalization in step S4 into the trained encoder-decoder neural network based on U-Net to obtain the breathing waveform. Obtain the breathing frequency detection value through the breathing waveform to complete the breathing monitoring based on ultra-wideband radio frequency signals.

[0012] Specifically, in step S2, the product of the respiratory rate components is:

[0013]

[0014] Among them, v i-1 v is the coefficient for the (i-1)th respiratory rate. i v is the coefficient for the (i+1)th respiratory rate. i+1 α is the coefficient for the i-th respiratory rate. p Let P be the distance interval index, p be the distance range index.

[0015] Specifically, in step S3, the method of removing the background reflection offset signal using the adjacent distance interval similarity transformation is as follows:

[0016] The signal matrix y(t,i) obtained in step S1 is modeled as the background reflection offset y b (t,i) and respiratory-induced signal y v The sum of (t, i), where t is the frame index and i is the distance interval index, is obtained by minimizing the y-axis on the I / Q constellation graph obtained in step S1. v (t,i) and y v The Euclidean distance and Γ of each point pair in (t, i+1) are used to remove the background reflection offset signal and obtain the signal triggered by the user's breathing.

[0017] Furthermore, the Euclidean distance and Γ are:

[0018]

[0019] in,(·) I and(·) Q These are the I / Q components of the corresponding parameters, where k is the scaling factor, θ is the rotation angle, and e is the rotation angle. j This is a unified expression for electromagnetic wave signals.

[0020] Specifically, in step S4, phase shift for:

[0021]

[0022] Where, θ n For the total phase shift, (·) I and(·) Q These are the I / Q components of the corresponding parameters, e j This is a unified expression for electromagnetic wave signals.

[0023] Specifically, in step S5, the loss function of the U-Net-based encoder-decoder neural network... for:

[0024]

[0025] Where λ is the weighting coefficient. Let latent feature consistency loss function be used. This is the mean squared error loss function.

[0026] Furthermore, the two breathing distance intervals are the two intervals i and j with the highest frequency product in step S2.

[0027] Secondly, embodiments of the present invention provide a respiratory monitoring system based on ultra-wideband radio frequency signals, comprising:

[0028] The sensing module uses pulse wireless ultra-wideband radar to send pulse signals to sense human respiration and the environment, and obtains a signal matrix. Each data point of the signal matrix along the time axis is placed into the complex plane to obtain the corresponding I / Q constellation diagram.

[0029] The positioning module performs a fast Fourier transform on each distance interval of the signal matrix obtained by the sensing module along the slow time direction, and then multiplies the spectral features of three adjacent consecutive intervals to locate the distance interval of the user's breathing in the signal matrix based on the maximum value within the breathing frequency range.

[0030] The elimination module models the signal matrix within the distance interval of the user's breathing obtained by the sensing module as the sum of the background reflection offset signal and the breathing-induced signal. Based on the I / Q constellation diagram obtained by the sensing module, the background reflection offset signal is removed by the adjacent distance interval similarity transformation method to obtain the breathing-induced signal.

[0031] The preprocessing module processes the respiratory evoked signal obtained by the elimination module, compensates for the phase shift caused by the position and orientation of the monitored person, and normalizes the amplitude of the respiratory evoked signal to obtain two normalized respiratory distance interval signals.

[0032] The monitoring module designs and trains an encoder-decoder neural network based on U-Net. The amplitude, real part I, and imaginary part Q of the signals from the two breathing distance intervals after normalization by the preprocessing module are fed into the trained encoder-decoder neural network based on U-Net to obtain the breathing waveform. The breathing frequency detection value is obtained from the breathing waveform, thus completing the breathing monitoring based on ultra-wideband radio frequency signals.

[0033] Thirdly, a computer device includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor, when executing the computer program, implements the steps of the above-described respiratory monitoring method based on ultra-wideband radio frequency signals.

[0034] Fourthly, embodiments of the present invention provide a computer-readable storage medium including a computer program, which, when executed by a processor, implements the steps of the above-described respiratory monitoring method based on ultra-wideband radio frequency signals.

[0035] Compared with the prior art, the present invention has at least the following beneficial effects:

[0036] A respiratory monitoring method based on ultra-wideband radio frequency signals utilizes commercial pulsed ultra-wideband (IR-UWB) radar for non-contact sensing of the human body. The received signal is decomposed into background reflection offset signal and respiratory-induced signal. The respiratory-induced signal is modeled, and an FFT is performed on each interval of the signal matrix. Multiplying the spectra of three adjacent intervals accurately locates the interval of human respiration within the signal matrix. The similarity transformation (ST) algorithm effectively eliminates background noise. Phase offset compensation and amplitude normalization eliminate the influence of irrelevant factors such as the monitored person's position and orientation. An encoder-decoder neural network is built and trained to output a more accurate respiratory waveform, reconstructing human respiration. This method integrates previous spatial modeling and neural network approaches, focusing on solving a key problem: how to perform noise cancellation and attenuation estimation on the sensed signal when the user's pose in space is unknown, thus eliminating interference from actual factors such as user distance, posture, and surrounding environment, and achieving more accurate user respiratory monitoring.

[0037] Furthermore, by performing an FFT on each interval of the signal matrix and then multiplying the spectra of three adjacent intervals, the interval in which human respiration occurs within the signal matrix can be accurately located.

[0038] Furthermore, similarity transformation algorithms can effectively eliminate background noise in the signal.

[0039] Furthermore, Euclidean distance and Γ are mainly used to calculate the parameters when the signal distance between two distance intervals in the similarity transformation algorithm is minimized, i.e., when they are most similar.

[0040] Furthermore, phase shift The settings can help estimate the impact of breathing-independent variables such as user posture and distance on breathing-induced signals.

[0041] Furthermore, the loss function of the encoder-decoder neural network based on U-Net. It can help train the network, extract similar parts (i.e., user breathing activity) in two distance intervals, and discard irrelevant parts (such as environmental interference).

[0042] Furthermore, the purpose of setting the two breathing distance intervals as the two intervals i and j with the highest frequency product in step S2 is: 1) A higher frequency product indicates that the user's breathing activity is more obvious in that distance interval; 2) Using the two intervals with the most obvious user breathing activity can better utilize similarity to extract user breathing activity and discard signal changes caused by irrelevant variables.

[0043] It is understood that the beneficial effects of the second to fourth aspects mentioned above can be found in the relevant descriptions in the first aspect mentioned above, and will not be repeated here.

[0044] In summary, this invention uses commercially available IR-UWB radar for non-contact monitoring of human respiration, which can be widely applied in the medical field, especially in scenarios where traditional contact-based respiratory monitoring is not suitable. The knowledge-data dual-driven approach eliminates the influence of environmental factors and is not limited to specific environments, avoiding the limitations of existing solutions and achieving high-precision reconstruction of human respiration.

[0045] The technical solution of the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. Attached Figure Description

[0046] Figure 1 This is a schematic diagram of the signal matrix;

[0047] Figure 2 This is a schematic diagram of the I / Q constellation corresponding to the signal;

[0048] Figure 3 This is a schematic diagram of the respiratory zone location;

[0049] Figure 4 A graph showing the difference between I / Q components before and after background noise removal using the ST algorithm;

[0050] Figure 5 This is a schematic diagram showing the angle between radar and human breathing.

[0051] Figure 6 The diagram shows the difference between I / Q components before and after phase shift compensation.

[0052] Figure 7 This is a diagram of the encoder-decoder neural network structure based on U-Net.

[0053] Figure 8 This is a schematic diagram of the scene setup during the experiment;

[0054] Figure 9 This is a diagram showing the respiratory monitoring results of the present invention under stable conditions;

[0055] Figure 10 This is a diagram showing the respiratory monitoring results of the present invention under dynamic conditions. Detailed Implementation

[0056] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0057] In the description of this invention, it should be understood that the terms "comprising" and "including" indicate the presence of the described features, integrals, steps, operations, elements and / or components, but do not exclude the presence or addition of one or more other features, integrals, steps, operations, elements, components and / or collections thereof.

[0058] It should also be understood that the terminology used in this specification is for the purpose of describing particular embodiments only and is not intended to limit the invention. As used in this 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.

[0059] It should also be further understood that the term "and / or" as used in this specification and the appended claims refers to any combination and all possible combinations of one or more of the associated listed items, and includes such combinations. For example, A and / or B can represent three cases: A alone, A and B simultaneously, and B alone. Additionally, the character " / " in this document generally indicates that the preceding and following objects have an "or" relationship.

[0060] It should be understood that although terms such as first, second, third, etc., may be used in the embodiments of the present invention to describe the preset range, these preset ranges should not be limited to these terms. These terms are only used to distinguish the preset ranges from one another. For example, without departing from the scope of the 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.

[0061] Depending on the context, the word "if" as used here can be interpreted as "when," "when," "in response to determination," or "in response to detection." Similarly, depending on the context, the phrase "if determination" or "if detection (of the stated condition or event)" can be interpreted as "when determination," "in response to determination," "when detection (of the stated condition or event)," or "in response to detection (of the stated condition or event)."

[0062] The accompanying drawings illustrate various structural schematic diagrams according to embodiments disclosed in this invention. These drawings are not to scale, and some details have been enlarged for clarity, and some details may have been omitted. The shapes of the various regions and layers shown in the drawings, as well as their relative sizes and positional relationships, are merely exemplary and may deviate from reality due to manufacturing tolerances or technical limitations. Furthermore, those skilled in the art can design regions / layers with different shapes, sizes, and relative positions as needed.

[0063] This invention provides a respiratory monitoring method based on ultra-wideband radio frequency signals. It acquires signal data returned by IR-UWB radar; performs FFT on the signal in each interval, and multiplies the spectra of three adjacent intervals, using the maximum coefficient within the corresponding frequency range as a metric to locate the interval in which the human is breathing; applies a similarity transformation method to the signal within the human breathing interval to eliminate background noise; compensates for phase shifts caused by the monitored subject's position and orientation, and normalizes the signal amplitude to eliminate the influence of irrelevant factors such as the monitored subject's position and orientation; further feeds the amplitude, I-component, and Q-component of the signal in two breathing intervals into a trained encoder-decoder neural network, which outputs a more accurate respiratory waveform, thus reconstructing human respiration. This invention first utilizes modeling to remove variables unrelated to user breathing, such as noise, user orientation, and distance. Secondly, it constructs a neural network model to further eliminate errors introduced by sudden variables, extracting accurate breathing waveforms. Commercial pulsed ultra-wideband (IR-UWB) radar is used to sense human breathing. Combining knowledge-driven and data-driven methods, the radar data is processed to eliminate the influence of irrelevant environmental factors, restoring the human breathing waveform. This achieves non-contact monitoring of human breathing and can be widely applied in the medical field.

[0064] This invention discloses a respiratory monitoring method based on ultra-wideband radio frequency signals, comprising the following steps:

[0065] S1. Use pulsed ultra-wideband (IR-UWB) radar to sense human breathing and the environment, and obtain the signal matrix y(t,i), where t is the frame index (which can be used as time) and i is the range interval index. For the signal of each range interval, it can be converted to the I / Q complex plane to obtain the corresponding I / Q constellation diagram.

[0066] The radar transmits frames at regular intervals and superimposes the received frames to form a signal matrix y(t,i);

[0067] Please see Figure 1t is the frame index (time), and i is the interval index, where each interval corresponds to the radar's range resolution. Changes caused by object movement can be retrieved along the time axis.

[0068] After I / Q downconversion, the demodulated signal can be converted into a constellation diagram on the complex plane;

[0069] Please see Figure 2 It shows a constellation diagram containing two intervals of user breathing, with each point representing the received signal at a given time point, which is a combination of background reflection (BGR) offset and breathing-induced changes (pure vector).

[0070] S2. Locate the user's breathing;

[0071] For each distance interval of the signal matrix y(t,i) obtained in step S1, perform a Fast Fourier Transform (FFT) along the time direction. Then, multiply the spectral features of three adjacent consecutive intervals. Use the maximum value of the product within the frequency range corresponding to breathing (human breathing is 12-20 times per minute) as a metric for locating breathing. At the distance corresponding to the user's breathing, due to the common frequency characteristics brought about by the user's breathing, the maximum value of this frequency product should be relatively obvious; while at other distances, since the noise space and the feature space are orthogonal, the product should approach zero. This method can accurately locate the distance interval of the user's breathing activity in the signal matrix.

[0072] The specific steps are as follows: Perform a Fourier transform on each interval to obtain the spectral characteristics:

[0073]

[0074] The spectral characteristics of three adjacent consecutive intervals are multiplied as follows:

[0075]

[0076] Where v is the coefficient of respiratory rate, i.e.

[0077] It is a component related to background noise, equal to:

[0078]

[0079] in, It is the frequency coefficient of the background noise.

[0080] Since the noises are independent of each other, the product of the coefficients is almost zero, that is...

[0081] The product of the components of respiratory rate is expressed as:

[0082]

[0083] Because the distance change Δd caused by the rise and fall of the user's chest is present in the signals received from adjacent distance intervals. v Very similar, therefore, v i-1 v i and v i+1 For the same frequency component 2Δd v f c / c has a non-negligible coefficient, therefore the product is non-zero. Using the maximum coefficient within the frequency range corresponding to breathing as a metric, it is possible to filter and locate the distance interval where user breathing exists in the radar signal matrix y(t,i).

[0084] Please see Figure 3 , Figure 3 (a) shows the spectrogram of intervals 19 to 21 from one experiment. All three intervals contain the user's breathing. Figure 3 (b) shows the product of the spectra of the three intervals. For comparison, it also shows the product of the spectra of intervals 5 to 7, which do not include user breathing. The maximum coefficient within the corresponding respiratory frequency range (human breathing is 12–20 breaths / minute) is then used as a metric to locate the interval in which human breathing occurs. Figure 3 (c) shows the effect of positioning. It can be seen that the product of intervals 19 to 21 is larger, while the product of other intervals is smaller, thus enabling the positioning of the distance interval where the user is breathing.

[0085] S3. Model the radar signal;

[0086] The signal matrix y(t,i) received by the radar in step S1 is modeled as the background reflection offset y b And breathing triggers signals y v Based on the similarity of changes caused by chest displacement in adjacent intervals, a knowledge-based noise reduction method, called similarity transformation (ST), is used to eliminate the background component y in the signal matrix y(y,i). b Receive breathing-triggered signals y v ;

[0087] The specific steps are as follows:

[0088] The signal matrix y(t,i) represents the background reflection offset y. b And breathing triggers signals y v The sum is:

[0089] y(t,i)=y b (t,i)+y v (t,i)

[0090] Among them, y b (t,i) is the background reflection (BGR) offset, including reflections of people, furniture, and accidental reflections, y v (t,i) is the respiratory trigger signal; human respiration is only related to y. v It is related to (t,i), and to the background reflection y. b (t,i) are unrelated.

[0091] Furthermore, breathing triggers signals y v (t,i) is modeled as follows:

[0092]

[0093] Where P is the number of reflection points, α p (i) is the energy coefficient of the p-th reflected signal, f c It is the carrier frequency, c is the speed of light, and d is the carrier frequency. p (t,i) is the distance from the sensor to the p-th reflection point, and the variable distance d p It consists of two parts, namely the average distance of the i-th interval. and due to changes in chest movement d v (t,i), that is,

[0094]

[0095] Observations revealed that the changes caused by chest displacement were very similar in adjacent intervals, namely:

[0096]

[0097] Where, Δd v (t,i)=d v (t+Δt,i)-d v (t,i). Therefore, after some rotation and scaling, respiration triggers the signal y. v (t,i) and y v (t, i+1) should match each other, that is:

[0098] y v (t,i)≈k·e jθ ·y v (t, i+1)

[0099] Where k is the scaling factor, k = ∑α p (i) / ∑α p (i+1), θ is the rotation angle, equal to:

[0100] θ=2πf c (d p (t,i)-d p(t,i+1)) / c

[0101] The rotation angle should be Where t0 is the start time, therefore, the rotation angle is independent of other factors and only depends on the initial phase, and this angle does not change with time.

[0102] Since the changes caused by chest displacement are very similar in adjacent intervals, this application proposes a similarity transformation (ST) method, the specific steps of which are as follows:

[0103] By using rotation and scaling transformations, the parameters that best fit the breathing motion within adjacent distance intervals are found, thereby estimating the errors caused by changes in the surrounding environment and human posture within different distance intervals. Considering that the trigger signal is the difference between the received signal y(t,i) and the background noise, we have:

[0104] [y(t,i)-y b [(t,i)]≈k·e jθ ·[y(t,i+1)-y b (t,i+1)]

[0105] To estimate the background component y b (i) and y b (i+1), scaling factor k, and rotation angle θ are minimized by calculating the Euclidean distance between the respiratory evoked signals in the (i+1)th and ith distance intervals on the I / Q plane; that is, by minimizing the respiratory evoked signal y on the obtained I / Q constellation diagram using the least squares method. v (t,i) and y v The sum of the Euclidean distances Γ for each pair of points (t, i+1), i.e.:

[0106]

[0107] in,(·) I and(·) Q These are the I / Q components of the corresponding parameters.

[0108] This method can be used to extract the background component y that is unrelated to breathing from the received signal. b (i) and y b (i+1) is used for effective estimation and removal to preserve the respiratory-inducing signal y. v .

[0109] Please see Figure 4 , Figure 4 This demonstrates the effectiveness of the ST algorithm in eliminating background noise. Figure 4 (a) and (b) were conducted at a distance of 1.3m, without any other interference. Figure 4(b)(c)(e)(f) were conducted at a distance of 0.9m. The monitored person was moving or there were moving pedestrians around. As can be seen from the images, this method can eliminate background noise interference in different distance ranges and obtain the signal caused by the user's breathing.

[0110] S4. The respiratory evoked signal y obtained in step S3 v Further processing is performed to compensate for phase shifts caused by the location and orientation of the monitored subject, and the amplitude is normalized. Since the distance of the user will cause different amplitude attenuations, introducing errors into respiratory signal monitoring, this invention does not estimate the specific signal attenuation, but instead normalizes the processed signal to avoid errors caused by different distances from the monitored subject. The specific steps are as follows:

[0111] First, estimate the phase error introduced by the position and orientation of the subject being tested;

[0112] The position of the monitored subject introduces an inherent initial phase θ0 = 4πf c The orientation of the monitored subject introduces the following phase error into d(t0,i) / f:

[0113] θ e (t)=4πf c [Δd v [(t,i)cos(φ)] / c

[0114] Please see Figure 5 φ is the angle between the breathing direction and the radial direction toward the radar receiver, and the total phase offset θ n (t)=θ0+θ e (t), in order to eliminate irrelevant factors, it is necessary to compensate for the phase shift θ. n (t)=θ0+θ e (t) and amplitude decay γ.

[0115] Since the location and orientation of the monitored object are unknown, phase shift and propagation attenuation cannot be directly predicted. Therefore, this invention proposes a reverse reasoning method to estimate the phase shift. The main idea is that the angle at which the signal change is most significant should be the phase error when the user is directly facing the radar, i.e., when φ = 0, the signal change is most significant.

[0116]

[0117] Please see Figure 6 , Figure 6 The difference before and after phase compensation is shown; the Q component signal is very indistinct before compensation. Figure 6 (a)), but it becomes more apparent after compensation. Figure 6(b)). By estimating this phase error, the irrelevant variable introduced by the subject's body orientation can be eliminated, making the signal independent of the subject's orientation.

[0118] In addition to phase shift, amplitude attenuation caused by path loss is also crucial. Here, instead of predicting the precise value of the attenuation parameter γ, the processed signal is normalized to [0,1] to mitigate the error introduced by distance-induced amplitude attenuation.

[0119] S5. To further recover the breathing waveform, an encoder-decoder neural network based on U-Net was designed. The trained neural network takes the amplitude, real part I and imaginary part Q components of the signal in two breathing distance intervals as input and outputs a more accurate breathing waveform.

[0120] Please see Figure 7 The fundamental idea is to learn the general change patterns in different intervals. The life movements in adjacent intervals show similar change trends, while background noise does not.

[0121] Data from two breathing distance intervals over time are input into a neural network to learn common features caused by the user's chest movements. The two distance intervals are i and j, which have the highest frequency product in step S2. The above data is then input into a dual-channel encoder to learn the latent features of the two intervals, and a latent feature consistency loss function is defined. for:

[0122]

[0123] Among them, h i and h j These are the potential characteristics of the corresponding signal;

[0124] After the encoder-decoder network, convolutional blocks are used and the results are fed into fully connected blocks to recover the breathing signal, employing the mean squared error loss function. for:

[0125]

[0126] Among them, s and These are the real signal and the predicted signal, respectively.

[0127] Total loss function for:

[0128]

[0129] Wherein, λ is the weighting coefficient, which is 5 in the experiment, or can be set to other values. The respiratory signal recovered by this network can be used to calculate specific parameters such as respiratory rate.

[0130] In another embodiment of the present invention, a respiratory monitoring system based on ultra-wideband radio frequency signals is provided. This system can be used to implement the above-mentioned respiratory monitoring method based on ultra-wideband radio frequency signals. Specifically, the respiratory monitoring system based on ultra-wideband radio frequency signals includes a sensing module, a positioning module, an elimination module, a preprocessing module, and a monitoring module.

[0131] The sensing module uses pulse wireless ultra-wideband radar to send pulse signals to sense human respiration and the environment, obtains a signal matrix, and puts each data point of the signal matrix along the time axis into the complex plane to obtain the corresponding I / Q constellation diagram.

[0132] The positioning module performs a fast Fourier transform on each distance interval of the signal matrix obtained by the sensing module along the slow time direction, and then multiplies the spectral features of three adjacent consecutive intervals to locate the distance interval of the user's breathing in the signal matrix based on the maximum value within the breathing frequency range.

[0133] The elimination module models the signal matrix within the distance interval of the user's breathing obtained by the sensing module as the sum of the background reflection offset signal and the breathing-induced signal. Based on the I / Q constellation diagram obtained by the sensing module, the background reflection offset signal is removed by the adjacent distance interval similarity transformation method to obtain the breathing-induced signal.

[0134] The preprocessing module processes the respiratory evoked signal obtained by the elimination module, compensates for the phase shift caused by the position and orientation of the monitored person, and normalizes the amplitude of the respiratory evoked signal to obtain two normalized respiratory distance interval signals.

[0135] The monitoring module designs and trains an encoder-decoder neural network based on U-Net. The amplitude, real part I, and imaginary part Q of the signals from the two breathing distance intervals after normalization by the preprocessing module are fed into the trained encoder-decoder neural network based on U-Net to obtain the breathing waveform. The breathing frequency detection value is obtained from the breathing waveform, thus completing the breathing monitoring based on ultra-wideband radio frequency signals.

[0136] In another embodiment of the present invention, a terminal device is provided, comprising a processor and a memory. The memory stores a computer program, the computer program including program instructions, and the processor executes the program instructions stored in the computer storage medium. The processor may be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. It is the computing and control core of the terminal, suitable for implementing one or more instructions, specifically suitable for loading and executing one or more instructions to achieve a corresponding method flow or corresponding function. The processor described in this embodiment of the present invention can be used for the operation of a respiratory monitoring method based on ultra-wideband radio frequency signals, including:

[0137] Pulse-based ultra-wideband radar transmits pulse signals to sense human respiration and the environment, obtaining a signal matrix. Each data point along the time axis of the signal matrix is ​​placed into the complex plane to obtain the corresponding I / Q constellation diagram. A Fast Fourier Transform is performed on each distance interval of the signal matrix along the time axis, and the spectral characteristics of three adjacent consecutive intervals are multiplied. The distance interval of the user's respiration within the signal matrix is ​​located based on the maximum value within the respiratory frequency range. The signal matrix within the user's respiration distance interval is modeled as the sum of background reflection offset signal and respiration-induced signal. Based on the I / Q constellation diagram, an adjacent distance interval similarity transformation method is used to remove... Background reflection offset signal is used to obtain respiratory induction signal; the respiratory induction signal is processed to compensate for phase offset caused by the position and orientation of the monitored person, and the amplitude of the respiratory induction signal is normalized to obtain two normalized respiratory distance interval signals; a U-Net-based encoder-decoder neural network is designed and trained, and the amplitude, real part I, and imaginary part Q of the two normalized respiratory distance interval signals are fed into the trained U-Net-based encoder-decoder neural network to obtain respiratory waveform. The respiratory frequency detection value is obtained through the respiratory waveform, thus completing respiratory monitoring based on ultra-wideband radio frequency signal.

[0138] In another embodiment of the present invention, a storage medium is also provided, specifically a computer-readable storage medium (memory). This computer-readable storage medium is a memory device in a terminal device used to store programs and data. It is understood that the computer-readable storage medium here can include both the built-in storage medium in the terminal device and extended storage media supported by the terminal device. The computer-readable storage medium provides storage space that stores the terminal's operating system. Furthermore, this storage space also stores one or more instructions suitable for loading and execution by a processor. These instructions can be one or more computer programs (including program code). It should be noted that the computer-readable storage medium here can be high-speed RAM or non-volatile memory, such as at least one disk storage device.

[0139] One or more instructions stored in a computer-readable storage medium can be loaded and executed by a processor to implement the corresponding steps of the respiratory monitoring method based on ultra-wideband radio frequency signals in the above embodiments; one or more instructions in the computer-readable storage medium are loaded and executed by the processor to perform the following steps:

[0140] Pulse-based ultra-wideband radar transmits pulse signals to sense human respiration and the environment, obtaining a signal matrix. Each data point along the time axis of the signal matrix is ​​placed into the complex plane to obtain the corresponding I / Q constellation diagram. A Fast Fourier Transform is performed on each distance interval of the signal matrix along the time axis, and the spectral characteristics of three adjacent consecutive intervals are multiplied. The distance interval of the user's respiration within the signal matrix is ​​located based on the maximum value within the respiratory frequency range. The signal matrix within the user's respiration distance interval is modeled as the sum of background reflection offset signal and respiration-induced signal. Based on the I / Q constellation diagram, an adjacent distance interval similarity transformation method is used to remove... Background reflection offset signal is used to obtain respiratory induction signal; the respiratory induction signal is processed to compensate for phase offset caused by the position and orientation of the monitored person, and the amplitude of the respiratory induction signal is normalized to obtain two normalized respiratory distance interval signals; a U-Net-based encoder-decoder neural network is designed and trained, and the amplitude, real part I, and imaginary part Q of the two normalized respiratory distance interval signals are fed into the trained U-Net-based encoder-decoder neural network to obtain respiratory waveform. The respiratory frequency detection value is obtained through the respiratory waveform, thus completing respiratory monitoring based on ultra-wideband radio frequency signal.

[0141] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. The components of the embodiments of the present invention described and shown in the accompanying drawings can generally be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of the present invention provided in the accompanying drawings is not intended to limit the scope of the claimed invention, but merely to illustrate selected embodiments of the invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without inventive effort are within the scope of protection of the present invention.

[0142] The respiratory monitoring method based on radio frequency signals of this invention was tested in the laboratory.

[0143] Please see Figure 8 , Figure 8 The experiment setup was shown, with the subject lying in bed and the invention used to monitor their breathing.

[0144] Please see Figure 9 , Figure 9 The demonstration shows the performance of this invention under relatively stable environmental conditions. The blue line represents the actual breathing signal, the gray dashed line represents the data output by this invention before training, and the black dashed line represents the predicted data after training. It can be seen that the untrained data is already very accurate, and the predicted data after training further improves the accuracy.

[0145] Please see Figure 10 , Figure 10 The diagram illustrates the performance of this invention after a person walks around the bed. Similarly, the blue line represents the actual breathing signal, the gray dashed line represents the untrained output data of this invention, and the black dashed line represents the predicted data after training. The results in the figure show that this invention performs well even in complex environments.

[0146] Experimental results show that, before training, the cosine similarity between the restored respiratory waveform and the actual respiratory waveform achieved by this invention is 0.9291. After training, the cosine similarity further improves to 0.9522. Considering the respiratory rate estimation error, the results show that before training, the respiratory rate estimation error of this invention is only 0.11 breaths / minute, and after training, the respiratory rate estimation error further decreases to 0.07 breaths / minute. This demonstrates that this invention can achieve high-precision non-contact monitoring of human respiration.

[0147] In summary, the present invention provides a respiratory monitoring method, medium, device, and system based on ultra-wideband radio frequency signals. It uses commercially available pulsed ultra-wideband (IR-UWB) radar to monitor human respiration without contact with the monitored person. The knowledge-data dual-driven approach effectively eliminates the influence of environmental factors and is not limited to specific environments, but can still achieve high-precision monitoring of human respiration in complex environments.

[0148] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product embodied on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0149] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0150] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0151] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0152] The above content is only for illustrating the technical concept of the present invention and should not be construed as limiting the scope of protection of the present invention. Any modifications made to the technical solution based on the technical concept proposed in this invention shall fall within the scope of protection of the claims of this invention.

Claims

1. A respiratory monitoring method based on ultra-wideband radio frequency signals, characterized in that, Includes the following steps: S1. Use pulse wireless ultra-wideband radar to send pulse signals to sense human breathing and the environment, obtain a signal matrix, and put each data of the signal matrix along the time axis into the complex plane to obtain the corresponding I / Q constellation diagram. S2. Perform a fast Fourier transform on each distance interval of the signal matrix obtained in step S1 along the time direction, and then multiply the spectral features of three adjacent consecutive intervals. Based on the maximum value within the respiratory frequency range, locate the distance interval of the user's breathing in the signal matrix. The product of the components of respiratory rate is: in, For the first The coefficient of respiratory rate, For the first The coefficient of respiratory rate, For the first The coefficient of respiratory rate, Let be the energy coefficient of the signal reflected by the p-th reflector. The number of all reflectors, Let p be the number of the p-th reflector. It is the index of the distance interval; S3. The signal matrix within the distance interval where the user's breathing is located, obtained in step S2, is modeled as the sum of the background reflection offset signal and the breathing-inducing signal. Based on the I / Q constellation diagram obtained in step S1, the background reflection offset signal is removed by the adjacent distance interval similarity transformation method to obtain the breathing-inducing signal. The signal matrix obtained in step S1 Modeled as background reflection offset And breathing triggers signals The sum of, It is the frame index. It is the index of the distance interval, obtained by minimizing the I / Q constellation diagram in step S1. and Euclidean distance of each pair of points and Remove background reflection offset signal, obtain signal triggered by user breathing, Euclidean distance and for: in, and These are the I / Q components of the corresponding parameters. For scaling factor, The rotation angle; S4. Process the respiratory evoked signal obtained in step S3 to compensate for the phase shift caused by the position and direction of the monitored person, and normalize the amplitude of the respiratory evoked signal to obtain two normalized respiratory distance interval signals. S5. Design and train an encoder-decoder neural network based on U-Net. Input the amplitude, real part I and imaginary part Q components of the two breathing distance interval signals after normalization in step S4 into the trained encoder-decoder neural network based on U-Net to obtain the breathing waveform. Obtain the breathing frequency detection value through the breathing waveform to complete the breathing monitoring based on ultra-wideband radio frequency signals.

2. The respiratory monitoring method based on ultra-wideband radio frequency signals according to claim 1, characterized in that, In step S4, phase shift for: in, This represents the total phase shift. and These are the I / Q components of the corresponding parameters. This is a unified expression for electromagnetic wave signals. A signal matrix triggered by respiration.

3. The respiratory monitoring method based on ultra-wideband radio frequency signals according to claim 1, characterized in that, In step S5, the loss function of the U-Net-based encoder-decoder neural network is... for: in, These are the weighting coefficients. Let latent feature consistency loss function be used. This is the mean squared error loss function.

4. The respiratory monitoring method based on ultra-wideband radio frequency signals according to claim 3, characterized in that, The two breathing distance intervals are the two intervals with the highest frequency product in step S2. and .

5. A respiratory monitoring system based on ultra-wideband radio frequency signals, characterized in that, include: The sensing module uses pulse wireless ultra-wideband radar to send pulse signals to sense human respiration and the environment, and obtains a signal matrix. Each data point of the signal matrix along the time axis is placed into the complex plane to obtain the corresponding I / Q constellation diagram. The positioning module performs a fast Fourier transform on each distance interval of the signal matrix obtained by the sensing module along the slow time direction, and then multiplies the spectral features of three adjacent consecutive intervals to locate the distance interval of the user's breathing in the signal matrix based on the maximum value within the breathing frequency range. The product of the components of respiratory rate is: in, For the first The coefficient of respiratory rate, For the first The coefficient of respiratory rate, For the first The coefficient of respiratory rate, Let be the energy coefficient of the signal reflected by the p-th reflector. The number of all reflectors, Let p be the number of the p-th reflector. It is the index of the distance interval; The elimination module models the signal matrix within the distance interval of the user's breathing obtained by the sensing module as the sum of the background reflection offset signal and the breathing-induced signal. Based on the I / Q constellation diagram obtained by the sensing module, the background reflection offset signal is removed by the adjacent distance interval similarity transformation method to obtain the breathing-induced signal. The signal matrix obtained in step S1 Modeled as background reflection offset And breathing triggers signals The sum of, It is the frame index. It is the index of the distance interval, obtained by minimizing the I / Q constellation diagram in step S1. and Euclidean distance of each pair of points and Remove background reflection offset signal, obtain signal triggered by user breathing, Euclidean distance and for: in, and These are the I / Q components of the corresponding parameters. For scaling factor, The rotation angle is specified in the preprocessing module. The preprocessing module processes the respiratory evoked signal obtained by the elimination module, compensates for the phase shift caused by the position and orientation of the monitored person, and normalizes the amplitude of the respiratory evoked signal to obtain two normalized respiratory distance interval signals. The monitoring module designs and trains an encoder-decoder neural network based on U-Net. The amplitude, real part I, and imaginary part Q of the signals from the two breathing distance intervals after normalization by the preprocessing module are fed into the trained encoder-decoder neural network based on U-Net to obtain the breathing waveform. The breathing frequency detection value is obtained from the breathing waveform, thus completing the breathing monitoring based on ultra-wideband radio frequency signals.

6. A computer-readable storage medium for storing one or more programs, characterized in that, The one or more programs include instructions that, when executed by a computing device, cause the computing device to perform any of the methods according to claims 1 to 4.

7. A computing device, characterized in that, include: One or more processors, a memory, and one or more programs, wherein the one or more programs are stored in the memory and configured to be executed by the one or more processors, the one or more programs including instructions for performing any of the methods according to claims 1 to 4.

Images