Respiration monitoring method and device based on WiFi signal

Abstract

The invention discloses a breath monitoring method and device based on a WiFi signal. The method comprises the steps that the WiFi signal reflected by the chest of a human body is collected to obtain channel state information, namely original CSI; preprocessing the original CSI time sequence of each subcarrier and then extracting a respiration signal; screening the respiration signals of the subcarriers to obtain a respiration signal set of high-quality subcarriers, performing spectral clustering grouping according to the fluctuation direction, and fusing the groups into fused respiration signals with consistent fluctuation direction; based on the distribution trend of the instantaneous amplitudes of the different-frequency subcarriers in the fused respiratory signal at the same moment in the frequency domain, linear fitting is conducted on the fused respiratory signal in the frequency domain, and the inspiration phase and the expiration phase are judged according to the fitting slope; and correcting the recognized inspiration phase and expiration phase to obtain a phase-labeled respiration waveform. According to the invention, high monitoring precision and robustness can be realized, and application scenes such as clinical diagnosis and health management can be effectively supported.

Description

Technical Field

[0001] This invention belongs to the field of wireless sensing and medical monitoring technology, specifically relating to a respiratory monitoring method and device based on WiFi signals. Background Technology

[0002] Breathing is one of the most important vital signs in the human body. Its rate, waveform, and inspiratory-to-expiratory ratio not only directly reflect the health of the cardiopulmonary system but also effectively indicate various clinical characteristics such as infection risk, psychological stress levels, and sleep quality. Accurate respiratory monitoring has extremely high medical value in clinical practice and routine health monitoring.

[0003] In existing technologies, wearable sensors (such as breathing belts and chest patches) are mainly used for respiratory monitoring. Although these methods have high accuracy, they suffer from drawbacks such as discomfort when worn, poor compliance, and difficulty in long-term use. To overcome these shortcomings, in recent years, the academic community has proposed a non-contact respiratory monitoring method based on WiFi signals. The basic principle is that the human chest cavity causes slight changes in wireless signals during breathing, and respiratory information can be inferred by analyzing the received signals.

[0004] Currently, the mainstream WiFi-based respiratory monitoring methods include: (1) Monitoring methods based on Received Signal Strength Indicator (RSSI), such as UbiBreathe, infer respiration by monitoring macroscopic fluctuations in signal strength. They have the advantages of simple systems and good compatibility. However, due to the low resolution and poor anti-interference of RSSI, its signal-to-noise ratio drops significantly in complex environments, making it difficult to achieve stable and high-precision respiratory monitoring.

[0005] (2) Monitoring methods based on Channel State Information (CSI), such as Wi-Sleep and Wi-ResP. These methods utilize CSI to provide more granular amplitude or phase information than RSSI, enabling more accurate capture of respiratory signals, and are currently the focus of research.

[0006] Despite the great potential shown by CSI-based methods, existing technologies still have the following shortcomings: (1) Limited functional dimensions: Most existing methods can only estimate the respiratory rate, but cannot clearly distinguish the inspiratory and expiratory modes based on the waveform. This deficiency makes it impossible to further calculate key clinical medical indicators such as the inspiratory-expiratory ratio.

[0007] (2) Insufficient spatial robustness: The monitoring performance of existing solutions is highly dependent on the user's spatial location relative to the WiFi device. Users usually need to be restricted to specific signal-sensitive areas to achieve good results; otherwise, the performance will be significantly reduced.

[0008] (3) Insufficient signal processing accuracy: Existing methods mostly use traditional methods such as simple subcarrier averaging or principal component analysis (PCA), which fail to make full use of the high-quality subcarriers in CSI that are sensitive to breathing, resulting in high distortion of the extracted breathing waveform and loss of detailed features, making it difficult to meet the requirements of medical diagnosis for waveform fidelity.

[0009] Therefore, there is an urgent need for a WiFi breathing monitoring technology that can overcome the above-mentioned shortcomings and support high accuracy and robustness. Summary of the Invention

[0010] In view of the above, the purpose of this invention is to provide a respiratory monitoring method and device based on WiFi signals, so as to continuously capture respiratory dynamics and distinguish inhalation and exhalation phases non-contactly in the user's natural activities and complex home environment, so as to achieve high-precision and robust respiratory detection and provide data support for clinical diagnosis and health management.

[0011] To achieve the above-mentioned objectives, the present invention provides the following technical solution: In a first aspect, an embodiment of the present invention provides a respiratory monitoring method based on WiFi signals, comprising the following steps: Collect WiFi signals reflected from the human chest cavity to obtain channel state information, i.e., raw CSI; Respiratory signals are extracted after preprocessing the original CSI time series of each subcarrier; After filtering the breathing signals of each subcarrier to obtain a set of high-quality subcarrier breathing signals, the signals are then clustered according to the wave direction, and the groups are merged into a fused breathing signal with the same wave direction. Based on the distribution trend of the instantaneous amplitude of different frequency subcarriers at the same moment in the frequency domain in the fused respiratory signal, a linear fit is performed on it in the frequency domain, and the inspiratory phase and expiratory phase are determined based on the fitting slope. The identified inspiratory and expiratory phases are corrected to obtain phase-labeled respiratory waveforms.

[0012] Specifically, the step of preprocessing the original CSI time series of each subcarrier to extract the respiratory signal includes: The original CSI time series of each subcarrier is denoised using a filter to retain the low-frequency respiratory signal components in the frequency range of 0.16Hz-0.5Hz. Then, the low-frequency respiratory signal components are smoothed using a local weighted regression scatter smoothing algorithm to extract the respiratory signal that can reflect the breathing situation.

[0013] Specifically, the step of filtering the respiratory signals of each subcarrier to obtain a set of high-quality subcarrier respiratory signals includes: For each subcarrier Its spectral energy distribution is obtained by performing a fast Fourier transform. And calculate the breathing signal-to-noise ratio for each subcarrier. , is represented as: , in, Indicates the frequency of the subcarrier. Indicates the frequency range of the respiratory signal; Set adaptive threshold , Represents the breathing signal-to-noise ratio of all subcarriers. the median of Represents the breathing signal-to-noise ratio of all subcarriers. The mean absolute deviation Indicates the adjustable coefficient, when When reserving subcarriers, when Subcarriers are removed, and the breathing signals of all retained subcarriers constitute a set of high-quality subcarrier breathing signals.

[0014] Specifically, the spectral clustering grouping according to the wave direction includes: Calculate any two subcarriers in the set of high-quality subcarrier respiratory signals. The cosine similarity between them is used to construct a similarity matrix. , is represented as: , in, Representing the similarity matrix The Middle Line 1 Column elements, and Representing any two subcarriers CSI time series, superscript Indicates transpose; Based on similarity matrix The subcarriers were divided into two groups using a spectral clustering algorithm, corresponding to two opposite oscillation directions during exhalation: monotonically decreasing and monotonically increasing CSI amplitude.

[0015] Specifically, fusing the groups into a fused respiratory signal with consistent wave direction includes: The waveforms of one group of subcarriers obtained from spectral clustering are sign-inverted to ensure that the fluctuation directions of all subcarriers in both groups are consistent. Weighted fusion is then performed to obtain the fused breathing signal. , is represented as: , in, Indicates time, Represents any group The set of subcarriers, Indicates subcarriers in any group CSI time series, The weighted parameter represents the positive correlation between the respiratory signal-to-noise ratio and the magnitude of the signal-to-noise ratio.

[0016] Specifically, the linear fitting of the instantaneous amplitude distribution trend of different frequency subcarriers in the fused respiratory signal at the same moment in the frequency domain includes: The fused respiratory signal is divided into continuous short time windows in the time series. The instantaneous amplitude of each subcarrier at the corresponding time is recorded. The instantaneous amplitude is Gaussian smoothed in the frequency domain. The distribution of the smoothed instantaneous amplitude with frequency is linearly fitted by least squares method.

[0017] Specifically, the method of determining the inspiratory and expiratory phases based on the fitting slope includes: Based on the fitting slope in the linear fitting results, phase judgment is performed within the current time window. If the fitting slope is greater than 0, the time period is determined to be the expiratory phase; if the fitting slope is less than 0, the time period is determined to be the inspiratory phase. By continuously judging the fitting slope of all windows in the time series, the phase labeling sequence of the entire fusion respiratory signal is obtained.

[0018] Specifically, the correction of the identified inspiratory and expiratory phases includes: Temporal smoothing constraint: Check the phase labels of adjacent time windows. If frequent reversals occur and the interval is less than the physiologically reasonable threshold, a majority voting mechanism is used for correction. Global waveform sign adjustment: The main oscillation direction of the fused respiratory signal is compared with the identified phase label. If they are opposite, the sign of the fused respiratory signal waveform is flipped to ensure that the rising segment of the waveform corresponds to inhalation and the falling segment corresponds to exhalation. After correction, the final breathing waveform with consistent direction and smooth noise is obtained.

[0019] Specifically, the method further includes: calculating respiratory-related medical parameters based on the finally obtained phase-labeled respiratory waveform, including at least respiratory rate, inspiratory-to-expiratory ratio, tidal volume variability, and approximate entropy.

[0020] Secondly, embodiments of the present invention also provide a respiratory monitoring device based on WiFi signals, implemented using the above-mentioned respiratory monitoring method based on WiFi signals, including: a transmitting end, a receiving end, a CSI data acquisition module, a signal processing module, and a respiratory parameter calculation module; The transmitting end is used to continuously transmit WiFi signals; The receiving end is used to receive WiFi signals reflected by the human chest cavity; The CSI data acquisition module is used to extract channel state information, i.e., raw CSI, from the receiving end. The signal processing module is used to preprocess the original CSI time series of each subcarrier to extract respiratory signals; after filtering the respiratory signals of each subcarrier to obtain a set of high-quality subcarrier respiratory signals, it performs spectral clustering and grouping according to the fluctuation direction, and merges each group into a fused respiratory signal with the same fluctuation direction; based on the distribution trend of the instantaneous amplitude of different frequency subcarriers at the same moment in the frequency domain in the fused respiratory signal, it performs linear fitting in the frequency domain, and determines the inspiratory and expiratory phases based on the fitting slope; the identified inspiratory and expiratory phases are corrected to obtain the phase-labeled respiratory waveform; The respiratory parameter calculation module is used to calculate respiratory-related medical parameters based on the respiratory waveform with the final inspiratory and expiratory phase labels.

[0021] Compared with the prior art, the beneficial effects of the present invention include at least the following: (1) Complete functions: By introducing a phase identification and correction process based on frequency domain distribution trend fitting analysis, this invention achieves accurate differentiation between the inspiratory and expiratory phases in non-contact monitoring, thus breaking through the limitation of existing technologies that can only estimate respiratory rate and providing a technical basis for the calculation of respiratory-related medical parameters.

[0022] (2) Strong robustness: This invention effectively utilizes the frequency domain diversity of WiFi signals by using adaptive subcarrier selection based on respiratory signal-to-noise ratio and combining spectral clustering and direction unification fusion strategy based on cosine similarity. This overcomes the problem of decreased signal sensitivity caused by changes in user location and achieves reliable respiratory monitoring without precise positioning within a typical room scale.

[0023] (3) Improved accuracy: This invention constructs a multi-level signal purification process of preprocessing, quality screening and cluster fusion, which maximizes the aggregation of high-quality subcarrier information sensitive to breathing and effectively suppresses noise and interference, so that the reconstructed respiratory waveform is highly consistent with the real respiratory motion, and the Pearson correlation coefficient (PCC) can stably exceed 85%, which provides a guarantee for accurate clinical analysis. Attached Figure Description

[0024] To more clearly illustrate the technical solutions in the embodiments of the present 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 only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0025] Figure 1 This is a schematic flowchart of the respiratory monitoring method based on WiFi signals provided in an embodiment of the present invention; Figure 2 This is a schematic diagram of the signal propagation path in a WiFi respiratory monitoring scenario provided by an embodiment of the present invention; Figure 3 This is a schematic diagram of three CSI amplitude variation modes provided in the embodiments of the present invention; Figure 4 This is a schematic diagram illustrating the results of filtering subcarriers within the bandwidth based on breathing signal-to-noise ratio, provided in an embodiment of the present invention. Figure 5 This is a schematic diagram illustrating the result of grouping subcarriers using a similarity-based spectral clustering algorithm, as provided in an embodiment of the present invention. Figure 6 This is a schematic diagram of the final phase-labeled respiratory waveform obtained according to an embodiment of the present invention; Figure 7 This is a schematic diagram of the structure of a respiratory monitoring device based on WiFi signals provided in an embodiment of the present invention. Detailed Implementation

[0026] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the invention and do not limit the scope of protection of this invention.

[0027] The inventive concept of this invention is as follows: Addressing the problems of existing technologies, such as the inability to distinguish inspiratory and expiratory phases, location-limited monitoring results, and insufficient accuracy, this invention provides a respiratory monitoring method and device based on WiFi signals. First, the original CSI time series is preprocessed to extract effective respiratory components. Then, an innovative adaptive threshold based on the respiratory signal-to-noise ratio is used to filter a set of high-quality subcarrier respiratory signals. Through spectral clustering and directional unification fusion, the impact of user location changes on monitoring stability is effectively overcome. Finally, by analyzing the distribution trend of the instantaneous amplitude of different frequency subcarriers in the frequency domain and performing linear fitting, accurate distinction and labeling of the inspiratory and expiratory phases are achieved.

[0028] like Figure 1As shown in the figure, an embodiment provides a respiratory monitoring method based on WiFi signals, including the following steps: S1 collects WiFi signals reflected from the human chest cavity to obtain channel state information, i.e., raw CSI.

[0029] During wireless signal propagation, the periodic undulations of the chest cavity cause slight changes in the length of the reflection path, resulting in periodic fluctuations in the amplitude and phase of the received signal over time. In this embodiment, a human body reflection signal model based on multipath propagation is first established, which clearly describes the relationship between CSI and chest cavity movement, i.e., the fluctuation of CSI amplitude corresponds periodically to the chest cavity displacement.

[0030] In WiFi-based respiratory monitoring scenarios, such as Figure 2 As shown, the main components of the received signal are the direct path (LoS) and the human chest cavity reflection path (bre, including the transmission path TB and the reflection path BR). Based on the principle of electromagnetic wave propagation, the human body reflection signal model based on multipath propagation using channel state information (CSI) is represented as follows: , in, The complex representation of the channel frequency response (CSI) is used to represent the channel frequency response. and These represent the signal amplitude coefficient and propagation distance along the direct path, respectively. and These represent the signal amplitude coefficient and total propagation distance of the chest cavity reflex path, respectively. , Represents the imaginary unit. Indicates the subcarrier frequency. Representing the speed of light, in The complex terms with base α represent the phase terms of the signal during propagation; Its amplitude The expression is: , Among them, phase difference This refers to the phase difference between electromagnetic wave signals traveling along direct and reflected paths when they reach the receiving end, and the phase difference between the phase difference and the frequency of the signal itself. and their respective transmission distances and Related, that is , This represents a constant phase term that is independent of propagation.

[0031] In this embodiment, by constructing this model, the chest displacement during respiration was revealed. Quantitative impact on CSI (e.g.) Figure 2 , Maximum phase change .

[0032] Based on this, different subcarriers, due to their different carrier frequencies, exhibit varying sensitivities to thoracic reflexes. Some show monotonous fluctuations, while others are weaker or more chaotic (here, subcarriers refer to WiFi subcarriers, which are also the frequency domain units for CSI). Further analysis of CSI variation patterns, such as... Figure 3 As shown, there are three CSI amplitude variation patterns: Mode 1 (Case 1): When the phase difference lie in During the exhalation phase, the chest expands, and the path length... Increase, CSI amplitude Monotonically decreasing; Mode 2 (Case 2): When the phase difference lie in During the interval, exhalation causes Monotonically increasing; Mode 3 (Case 3): When the phase difference Cross At the boundary, the non-monotonicity of the cosine function causes irregular fluctuations in the CSI amplitude.

[0033] S2 extracts respiratory signals after preprocessing the original CSI time series of each subcarrier.

[0034] Signal preprocessing is a crucial step in ensuring the extraction of respiratory signals from the raw CSI. In this embodiment, a Savitzky-Golay filter is first used to denoise the raw CSI on each subcarrier, preserving low-frequency respiratory signal components within the 0.16Hz-0.5Hz frequency range (corresponding to a respiratory rate of 10-30 breaths / minute). Then, a local weighted regression scatter smoothing algorithm (LOWESS) is used to smooth the low-frequency respiratory signal components, extracting subtle changes caused by chest cavity movement. Finally, a preliminary respiratory signal reflecting the respiratory status is extracted from each subcarrier of the CSI.

[0035] S3: After filtering the respiratory signals of each subcarrier to obtain a set of high-quality subcarrier respiratory signals, the signals are then clustered according to the wave direction and the groups are merged into a fused respiratory signal with the same wave direction.

[0036] To overcome the impact of location on respiratory monitoring performance and improve waveform accuracy, this invention proposes a subcarrier selection and fusion strategy, which aims to optimize the coarse waveform of the respiratory signal obtained in the previous step to make it closer to the true value.

[0037] First, let's introduce the theoretical basis of this strategy.

[0038] The above theory reveals the fundamental reason why body position affects respiratory performance: for a given frequency The subcarrier, if the corresponding phase difference range It did not fall completely into the range or If the amplitude changes, it may be non-monotonic or fluctuate slightly (mode 3), making it difficult to capture effective respiratory signal patterns. This phenomenon means that CSI monitoring at a single frequency is extremely sensitive to location.

[0039] However, a WiFi system is itself a broadband signal system with bandwidth. It can provide subcarriers with multiple frequency distributions. Therefore, it is possible to maintain the user's location without changing the user's position (i.e., keeping the user's location constant). Without changing the subcarrier frequency, the phase difference range is "shifted". According to the formula above, when a frequency offset of is selected... When the subcarrier is , its phase change is: , This means that by selecting subcarriers of different frequencies, the phase range can be adjusted. It falls into a more favorable range (e.g., mode 1 or mode 2), thus avoiding the non-monotonic mode 3. In other words, the subcarrier diversity in the frequency domain can compensate for the limitations of spatial location, enabling robust breathing monitoring in different locations.

[0040] Calculations based on WiFi bandwidth standards (WiFi 7 can reach 160MHz or even 320MHz) show that as long as the system bandwidth... The path difference satisfies the following conditions: This ensures that, within a certain set of subcarriers, at least some subcarriers exhibit a monotonic CSI variation pattern (mode 1 or mode 2), thus allowing the extraction of a stable respiratory signal. This theory guarantees that, in real-world environments (room scale), The system is position-robust and does not require precise control of the relative position of the human body and the equipment.

[0041] Therefore, given that there are high-quality subcarriers reflecting breathing among all CSI subcarriers, it is necessary to select subcarriers and remove invalid data. To this end, in this embodiment, the Breathing-to-Noise Ratio (BNR) metric is used in the frequency domain to quantify subcarrier quality. The specific implementation steps are as follows.

[0042] S3.1, for each subcarrier The spectral energy distribution is obtained by performing a Fast Fourier Transform (FFT). And calculate the breathing signal-to-noise ratio for each subcarrier. , is represented as: , in, Indicates the frequency of the subcarrier. This indicates the frequency range of the respiratory signal, i.e. a =0.16Hz, b =0.5Hz.

[0043] For each subcarrier, the respiratory signal-to-noise ratio (SNR) of the CSI time series is calculated, and an adaptive threshold is set. , Represents the breathing signal-to-noise ratio of all subcarriers. the median of Represents the breathing signal-to-noise ratio of all subcarriers. The mean absolute deviation Indicates the adjustable coefficient. When When reserving subcarriers, when Subcarriers are removed, and the respiratory signals of all retained subcarriers constitute a high-quality subcarrier respiratory signal set. .like Figure 4 The image shows an example of the results of filtering subcarriers within the bandwidth based on breathing signal-to-noise ratio. This filtering step effectively removes interfering or breathing-related subcarriers, improving the signal-to-noise ratio of subsequent signal fusion.

[0044] S3.2, within the respiratory signal set of high-quality subcarriers, two opposing fluctuation patterns still exist: one where the amplitude decreases during exhalation (pattern 1), and the other where the amplitude increases during exhalation (pattern 2). Direct synthesis would result in the effective signals canceling each other out. Cosine similarity, a metric for measuring the directionality of a time series, can be used to distinguish between these two patterns. Therefore, in this embodiment, a similarity-based spectral clustering algorithm is employed to automatically differentiate subcarriers with different trends.

[0045] For any two subcarriers Calculate the cosine similarity between them to construct a similarity matrix. , is represented as: , in, Representing the similarity matrix The Middle Line 1 Column elements, and Representing any two subcarriers CSI time series, superscript This indicates transpose.

[0046] Based on similarity matrix The subcarriers are divided into two groups using a spectral clustering algorithm. , These correspond to two different trends. For example... Figure 5 The diagram shows the results before and after spectral clustering.

[0047] S3.3, the waveforms of one group of subcarriers obtained from spectral clustering are sign-reversed to ensure that the fluctuation directions of all subcarriers in both groups are consistent, and then weighted and fused to obtain the fused breathing signal. , is represented as: ,

[0048] in, Indicates time, Represents any group The set of subcarriers, Indicates subcarriers in any group CSI time series, The weighted parameter represents the positive correlation between the respiratory signal-to-noise ratio and the magnitude of the signal-to-noise ratio.

[0049] Through these steps, a waveform that closely resembles real breathing, but does not yet correspond to specific inhalation and exhalation, is obtained.

[0050] S4. Based on the distribution trend of the instantaneous amplitude of different frequency subcarriers in the frequency domain in the fused respiratory signal, the fused respiratory signal is linearly fitted, and the inspiratory phase and expiratory phase are determined according to the fitting slope.

[0051] While the waveform of a fused respiratory signal reflects the respiratory rhythm, it's impossible to directly determine which segment represents inspiration and which represents expiration. In this embodiment, a phase recognition method based on frequency trends is proposed. The principle is that due to the CSI phase difference... As the frequency changes linearly, the instantaneous amplitudes of subcarriers of different frequencies exhibit a regular gradient trend at the same moment. The effect of chest cavity expansion (inhalation) on CSI amplitude is consistent with the effect of a lower subcarrier frequency, and conversely, the effect of exhalation on CSI amplitude is consistent with the effect of a higher subcarrier frequency. To achieve the differentiation between inspiratory and expiratory phases, the specific implementation steps are as follows.

[0052] S4.1, the fused respiratory signal is divided into continuous short time windows (approximately 0.2-0.5 s) in time series, and the instantaneous amplitude of each subcarrier at the corresponding time is recorded to obtain... ,in For the first The frequency of each subcarrier The corresponding instantaneous amplitude, This represents the number of subcarriers.

[0053] S4.2, Instantaneous amplitude of each subcarrier Gaussian smoothing is applied along the frequency direction to suppress random noise and preserve the overall trend, resulting in the smoothed instantaneous amplitude distribution as a function of frequency. , This represents a Gaussian smoothing kernel that performs a weighted average of the amplitude values ​​between subcarriers of different frequencies.

[0054] S4.3, for Perform a least squares linear fit to obtain the fitting result. .

[0055] S4.4, Based on the fitting slope in the linear fitting results, perform phase determination within the current time window. If the fitting slope... If the fitted slope is [value missing], then that time period is determined to be the expiratory phase. If the slope is zero, then that time period is determined to be the inspiratory phase. Combining theory and practice, the case where the slope is 0 can be disregarded here.

[0056] By continuously judging the fitting slope of all windows in the time series, the phase labeling sequence of the entire fused respiratory signal is obtained.

[0057] S5 corrects the identified inspiratory and expiratory phases to obtain phase-labeled respiratory waveforms.

[0058] In actual data, due to high noise in some subcarriers, the trend slope of individual windows may vary. Random phase reversals may occur. To ensure the continuity and physiological rationality of phase recognition, two correction mechanisms are introduced: (1) Temporal smoothing constraint: Check the phase labels of adjacent time windows. If frequent reversals occur and the interval is less than the physiologically reasonable threshold (i.e., the breathing cycle should be greater than 1s and the reversal interval should be less than 0.3s), then correct it according to the majority voting method of the forward and backward trends.

[0059] (2) Global adjustment of waveform symbol: The main oscillation direction (rising / falling) of the fusion respiratory signal is compared with the identified phase label. If they are opposite, the waveform of the fusion respiratory signal is reversed to ensure that the rising segment of the waveform corresponds to inhalation and the falling segment corresponds to exhalation.

[0060] After this correction step, a phase-labeled respiratory waveform with consistent direction and smoothed noise is finally obtained. The result is as follows: Figure 6 As shown, the orange waveform represents the actual result measured by the breathing zone, while the blue waveform represents the monitoring result of this invention, which has been corrected for direction.

[0061] Furthermore, in practical clinical diagnosis and health management applications, it also supports the calculation of respiratory-related medical parameters, including respiratory rate, inspiratory-expiratory ratio, tidal volume variability, and approximate entropy, based on the final phase-annotated respiratory waveform.

[0062] In summary, the respiratory monitoring method based on WiFi signals provided by this invention systematically solves the core difficulties of existing non-contact respiratory monitoring in terms of phase differentiation, location dependence, and signal accuracy through a multi-level processing flow of signal preprocessing, intelligent subcarrier screening and fusion, and phase recognition based on frequency trends. It significantly improves waveform fidelity and robustness of monitoring results, providing a reliable monitoring tool for clinical diagnosis.

[0063] Based on the same inventive concept, such as Figure 7 As shown, this embodiment of the invention also provides a respiratory monitoring device 700 based on WiFi signals, including: a transmitter 710, a receiver 720, a CSI data acquisition module 730, a signal processing module 740, and a respiratory parameter calculation module 750.

[0064] The transmitter 710 is used to continuously transmit WiFi signals.

[0065] The receiver 720 is used to receive WiFi signals reflected from the human chest cavity.

[0066] The CSI data acquisition module 730 is used to extract channel state information, i.e., raw CSI, from the receiver.

[0067] The signal processing module 740 is used to preprocess the original CSI time series of each subcarrier and extract the respiratory signal; after filtering the respiratory signals of each subcarrier to obtain a set of high-quality subcarrier respiratory signals, it performs spectral clustering and grouping according to the fluctuation direction, and merges each group into a fused respiratory signal with the same fluctuation direction; based on the distribution trend of the instantaneous amplitude of different frequency subcarriers at the same moment in the frequency domain in the fused respiratory signal, it performs linear fitting in the frequency domain, and determines the inspiratory phase and expiratory phase according to the fitting slope; the identified inspiratory phase and expiratory phase are corrected to obtain the phase-labeled respiratory waveform.

[0068] The respiratory parameter calculation module 750 is used to calculate respiratory-related medical parameters based on the respiratory waveform with the final inspiratory and expiratory phase labels.

[0069] It should be noted that the respiratory monitoring device based on WiFi signals provided in the above embodiments belongs to the same inventive concept as the respiratory monitoring method based on WiFi signals. For details of its implementation process, please refer to the embodiments of the respiratory monitoring method based on WiFi signals, which will not be repeated here.

[0070] The specific embodiments described above illustrate the technical solution and beneficial effects of the present invention in detail. It should be understood that the above description is only the most preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, additions, and equivalent substitutions made within the scope of the principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A respiratory monitoring method based on WiFi signals, characterized in that, Includes the following steps: Collect WiFi signals reflected from the human chest cavity to obtain channel state information, i.e., raw CSI; Respiratory signals are extracted after preprocessing the original CSI time series of each subcarrier; After filtering the breathing signals of each subcarrier to obtain a set of high-quality subcarrier breathing signals, the signals are then clustered according to the wave direction, and the groups are merged into a fused breathing signal with the same wave direction. Based on the distribution trend of the instantaneous amplitude of different frequency subcarriers at the same moment in the frequency domain in the fused respiratory signal, a linear fit is performed on it in the frequency domain, and the inspiratory phase and expiratory phase are determined based on the fitting slope. The identified inspiratory and expiratory phases are corrected to obtain phase-labeled respiratory waveforms.

2. The respiratory monitoring method based on WiFi signals according to claim 1, characterized in that, The step of preprocessing the original CSI time series of each subcarrier to extract the respiratory signal includes: The original CSI time series of each subcarrier is denoised using a filter to retain the low-frequency respiratory signal components in the frequency range of 0.16Hz-0.5Hz. Then, the low-frequency respiratory signal components are smoothed using a local weighted regression scatter smoothing algorithm to extract the respiratory signal that can reflect the breathing situation.

3. The respiratory monitoring method based on WiFi signals according to claim 1, characterized in that, The process of filtering the respiratory signals of each subcarrier to obtain a set of high-quality subcarrier respiratory signals includes: For each subcarrier Its spectral energy distribution is obtained by performing a fast Fourier transform. And calculate the breathing signal-to-noise ratio for each subcarrier. , is represented as: ， in, Indicates the frequency of the subcarrier. Indicates the frequency range of the respiratory signal; Set adaptive threshold , Represents the breathing signal-to-noise ratio of all subcarriers. the median of Represents the breathing signal-to-noise ratio of all subcarriers. The mean absolute deviation Indicates the adjustable coefficient, when When reserving subcarriers, when Subcarriers are removed, and the breathing signals of all retained subcarriers constitute a set of high-quality subcarrier breathing signals.

4. The respiratory monitoring method based on WiFi signals according to claim 1 or 3, characterized in that, The spectral clustering grouping based on wave direction includes: Calculate any two subcarriers in the set of high-quality subcarrier respiratory signals. The cosine similarity between them is used to construct a similarity matrix. , is represented as: ， in, Representing the similarity matrix The Middle Line 1 Column elements, and Representing any two subcarriers CSI time series, superscript Indicates transpose; Based on similarity matrix The subcarriers were divided into two groups using a spectral clustering algorithm, corresponding to two opposite oscillation directions during exhalation: monotonically decreasing and monotonically increasing CSI amplitude.

5. The respiratory monitoring method based on WiFi signals according to claim 4, characterized in that, The process of fusing the various groups into a fused respiratory signal with consistent wave direction includes: The waveforms of one group of subcarriers obtained from spectral clustering are sign-inverted to ensure that the fluctuation directions of all subcarriers in both groups are consistent. Weighted fusion is then performed to obtain the fused breathing signal. , is represented as: ， in, Indicates time, Represents any group The set of subcarriers, Indicates subcarriers in any group CSI time series, The weighted parameter represents the positive correlation between the respiratory signal-to-noise ratio and the magnitude of the signal-to-noise ratio.

6. The respiratory monitoring method based on WiFi signals according to claim 1, characterized in that, The method of linearly fitting the instantaneous amplitude distribution trend of different frequency subcarriers in the fused respiratory signal at the same moment in the frequency domain includes: The fused respiratory signal is divided into continuous short time windows in the time series. The instantaneous amplitude of each subcarrier at the corresponding time is recorded. The instantaneous amplitude is Gaussian smoothed in the frequency domain. The distribution of the smoothed instantaneous amplitude with frequency is linearly fitted by least squares method.

7. The respiratory monitoring method based on WiFi signals according to claim 1 or 6, characterized in that, The method of determining the inspiratory and expiratory phases based on the fitted slope includes: Based on the fitting slope in the linear fitting results, phase judgment is performed within the current time window. If the fitting slope is greater than 0, the time period is determined to be the expiratory phase; if the fitting slope is less than 0, the time period is determined to be the inspiratory phase. By continuously judging the fitting slope of all windows in the time series, the phase labeling sequence of the entire fusion respiratory signal is obtained.

8. The respiratory monitoring method based on WiFi signals according to claim 1, characterized in that, The correction of the identified inspiratory and expiratory phases includes: Temporal smoothing constraint: Check the phase labels of adjacent time windows. If frequent reversals occur and the interval is less than the physiologically reasonable threshold, a majority voting mechanism is used for correction. Global waveform sign adjustment: The main oscillation direction of the fused respiratory signal is compared with the identified phase label. If they are opposite, the sign of the fused respiratory signal waveform is flipped to ensure that the rising segment of the waveform corresponds to inhalation and the falling segment corresponds to exhalation. After correction, the final breathing waveform with consistent direction and smooth noise is obtained.

9. The respiratory monitoring method based on WiFi signals according to claim 1, characterized in that, The method further includes: calculating respiratory-related medical parameters based on the finally obtained phase-labeled respiratory waveform, including at least respiratory rate, inspiratory-expiratory ratio, tidal volume variability, and approximate entropy.

10. A respiratory monitoring device based on WiFi signals, implemented using the respiratory monitoring method based on WiFi signals according to any one of claims 1 to 9, characterized in that, include: Transmitter, receiver, CSI data acquisition module, signal processing module, and respiratory parameter calculation module; The transmitting end is used to continuously transmit WiFi signals; The receiving end is used to receive WiFi signals reflected by the human chest cavity; The CSI data acquisition module is used to extract channel state information, i.e., raw CSI, from the receiving end. The signal processing module is used to preprocess the original CSI time series of each subcarrier and extract the respiratory signal; after filtering the respiratory signals of each subcarrier to obtain a set of high-quality subcarrier respiratory signals, the module performs spectral clustering and grouping according to the fluctuation direction, and merges each group into a fused respiratory signal with the same fluctuation direction. Based on the frequency domain distribution trend of the instantaneous amplitude of different frequency subcarriers at the same moment in the fused respiratory signal, the fused respiratory signal is linearly fitted, and the inspiratory and expiratory phases are determined according to the fitting slope; the identified inspiratory and expiratory phases are corrected to obtain the phase-labeled respiratory waveform. The respiratory parameter calculation module is used to calculate respiratory-related medical parameters based on the respiratory waveform with the final inspiratory and expiratory phase labels.

Images