Dual-mode household respiration monitoring device integrating radar and WiFi channel state information

Abstract

The invention discloses a radar and WiFi channel state information fused dual-mode household respiration monitoring device which comprises a radar signal collecting and processing module used for detecting a human body respiration signal through a millimeter wave FMCW radar and processing an I / Q orthogonal echo signal obtained by the FMCW radar to obtain a respiration signal; the WiFi signal acquisition and processing module is used for acquiring a CSI signal containing respiratory activity information through a WiFi wireless network card and calculating a CSI ratio of a receiving channel so as to remove time-varying noise and further extract a respiratory signal; the data communication module is used for constructing a networking system based on a millimeter wave FMCW radar and WiFi through different network cable transmission modes; and the breathing information fusion module is used for dynamically fusing the two breathing signals. Starting from multi-sensor respiration signal extraction, accurate detection of personnel respiration signals in a home indoor environment is realized through a fusion algorithm, and the detection method is effective, feasible and reliable in performance.

Description

Technical Field

[0001] The present invention belongs to the technical field of home health monitoring, and in particular to a home respiratory monitoring device integrating radar and WiFi channel status information in dual modes. Background Art

[0002] Establishing a health monitoring and management mechanism is crucial. By implementing home health monitoring, health problems can be detected and intervened in a timely manner, thereby promoting early diagnosis and treatment.

[0003] Breathing is one of the most basic physiological activities of the human body and can directly reflect a person's health status. Home respiratory monitoring plays an important role in early warning, chronic disease management, sleep improvement, reducing medical burdens, improving quality of life, and protecting special populations. It is a key tool in modern health management. Traditional home respiratory monitoring methods mostly use contact or wearable methods, which are uncomfortable to wear and require frequent charging. They cannot provide full coverage of respiratory activity in complex home settings throughout the day and in all scenarios.

[0004] In recent years, research on non-contact respiratory monitoring technologies has rapidly increased. Common non-contact respiratory monitoring methods include audio, video, infrared, and radio frequency sensors. Among these technologies, radar- or WiFi-based radio frequency sensing has become a hot topic in research and application due to its advantages of fine-grained sensing, strong real-time performance, wide monitoring range, and good privacy protection.

[0005] However, home scenarios are complex and diverse, and channel state information from both radar and Wi-Fi may be flawed in certain scenarios. A single RF sensor struggles to effectively cover a home environment. Overall, methods using only radar or Wi-Fi channel state information alone are unable to meet the needs of 24 / 7 respiratory monitoring in complex home scenarios. Therefore, fully integrating radar and Wi-Fi channel state information from these two RF modalities effectively improves the performance of home respiratory monitoring and is the preferred solution for achieving 24 / 7 respiratory monitoring at home. Summary of the Invention

[0006] The purpose of the present invention is to address the problems existing in the above-mentioned prior art and provide a dual-modal home respiratory monitoring device that integrates radar and WiFi channel status information.

[0007] The technical solution to achieve the purpose of the present invention is: a dual-modal home respiratory monitoring device that integrates radar and WiFi channel status information, the device including a radar signal acquisition and processing module, a WiFi signal acquisition and processing module, a data communication module, and a respiratory information fusion module;

[0008] The radar signal acquisition and processing module is used to detect the human breathing signal through the millimeter wave FMCW radar, and process the I / Q orthogonal echo signal obtained by the FMCW radar to obtain the breathing signal;

[0009] The WiFi signal acquisition and processing module is used to obtain the CSI signal containing respiratory activity information through the WiFi wireless network card, calculate the CSI ratio of the receiving channel to remove time-varying noise, and then extract the respiratory signal;

[0010] The data communication module is used to build a networking system based on millimeter wave FMCW radar and WiFi through different network cable transmission modes;

[0011] The respiratory information fusion module is used to dynamically fuse the respiratory signals output by the radar signal acquisition and processing module and the WiFi signal acquisition and processing module.

[0012] Furthermore, the radar signal acquisition and processing module is used to detect the human breathing signal through the millimeter wave FMCW radar and process the I / Q orthogonal echo signal obtained by the FMCW radar to obtain the breathing signal, specifically including:

[0013] Step 1-1 uses a millimeter-wave FMCW radar to collect and extract in-phase and quadrature signals, or I / Q signals, within multiple range gates. Specifically, the FMCW radar's transmit synthesizer generates a linear frequency-modulated continuous wave signal, which is reflected when it encounters a human body. The millimeter-wave FMCW radar's quadrature receiver then captures the echo signal and performs quadrature mixing with the transmitted signal. The high-frequency portion is filtered out by a low-pass filter to produce an intermediate frequency (IF) signal. The IF signal is then sampled by an ADC and converted to digital. Range information is obtained through a range FFT, and phase data within the range gate is selected and stored.

[0014] Step 1-2: preprocessing the in-phase and quadrature signals, i.e., I / Q signals, within the multiple range gates to eliminate DC bias in the signals, demodulate target information, and filter out interference noise;

[0015] Step 1-3, reconstructing the multiple range gate signals pre-processed in step 1-2 using a method based on non-negative matrix factorization to eliminate random body motion interference;

[0016] Step 1-4: For the signals of the multiple range gates obtained after reconstruction in step 1-3, a range gate is selected based on the breathing quality index to output a breathing signal.

[0017] Furthermore, the WiFi signal acquisition and processing module is used to obtain a CSI signal containing respiratory activity information through a WiFi wireless network card, calculate the CSI ratio of the receiving channel to remove time-varying noise, and then extract the respiratory signal, specifically including:

[0018] Step 2-1: Obtaining an original CSI signal containing respiratory activity information through a WiFi wireless network card. Specifically, the WiFi wireless network card, in a multi-transmitter and multi-receiver configuration, captures a CSI data packet containing respiratory activity information. The CSI data packet is a complex matrix with dimensions N×M×P, where N is the number of transmitting antennas, M is the number of receiving antennas, and P is the number of subcarrier channels captured by the WiFi wireless network card.

[0019] Step 2-2: Calculate the CSI ratio of the two receiving antennas to obtain a CSI ratio data packet, thereby removing time-varying noise.

[0020] For each CSI data packet corresponding to a timestamp: For each transmit antenna, calculate the CSI ratios (CSI Ratios) for all possible receive antenna pairs, generating W rows × 1 column of Ratio values. This is then used to form a W × Q CSIRatio matrix, where W is the total number of subcarriers for all possible receive antenna pairs, and Q is the number of CSI Ratio data packets.

[0021] Step 2-3: Perform correction based on prior information, phase static component compensation, and phase jump compensation on the CSI Ratio data packet output in step 2-2, and output an updated W×Q1 CSI Ratio matrix, where Q1 is the number of updated CSI Ratio data packets.

[0022] Step 2-4: Divide the W×Q1 CSI Ratio matrix output from step 2-3 into w channels, each of which includes subcarriers; for each channel The subcarriers are combined based on the maximum ratio combining and principal component analysis method (RQI-MRC-PCA) of the RQI. Then, w channels are selected based on the respiratory quality index (RQI) to generate the final respiratory waveform.

[0023] Furthermore, the breathing information fusion module is used to dynamically fuse the breathing signals output by the radar signal acquisition and processing module and the WiFi signal acquisition and processing module, specifically including:

[0024] Step 4-1, extracting the respiratory frequency that changes over time from the respiratory signals output by the radar signal acquisition and processing module and the WiFi signal acquisition and processing module, specifically includes:

[0025] (1) Wavelet transform to extract time-frequency information: Perform wavelet transform on the respiratory signal x(t) to obtain the values of the scale a and time b parameters and the time-frequency representation W f (a,b) expression:

[0026]

[0027] Among them, a is the scale parameter, which is inversely proportional to the frequency; b is the time parameter, corresponding to the time axis position; ψ(*) is the mother wavelet function, which is used for local frequency analysis. is the conjugate function of ψ(*); the wavelet transform decomposes the signal into components of different scales, and obtains the time-scale joint distribution or the time-frequency representation W f (a, b), reflects the energy distribution of the signal at time b and scale a;

[0028] (2) Calculate the candidate instantaneous frequency: For W f (a,b)≠0, the candidate instantaneous frequency ω(a,b) is calculated by the phase derivative:

[0029]

[0030] Among them, the partial derivative It represents the rate of change with respect to time b, and the symbol i represents the imaginary unit;

[0031] (3) Synchronous compression transformation: transform the time-frequency representation W f (a, b) Redistribute along the frequency axis to generate a high-resolution time-frequency distribution T f (ω,b):

[0032]

[0033] Among them, A(b) is the effective scale set, satisfying |W f (a,b)|>∈, ∈ is the threshold used to suppress noise; δ(·) is the Dirac function that redistributes energy to the estimated instantaneous frequency ω;

[0034] (4) Extract instantaneous respiratory frequency: For each time b, in the time-frequency distribution T f Find the frequency corresponding to the energy peak in (ω,b) as the instantaneous breathing frequency f r (t):

[0035] f r (t) = arg max (T f (ω,b))

[0036] Among them, arg max(T f (ω,b)) represents the independent variable ω corresponding to the maximum value of energy intensity;

[0037] Step 4-2: For the respiratory signals output by the radar signal acquisition and processing module and the WiFi signal acquisition and processing module, the signal quality of the sensors is comprehensively evaluated using three independent indicators: respiratory quality index, respiratory frequency, and signal fluctuation standard deviation. Specifically, the following indicators are used:

[0038] Step 4-2-1, calculate the sensor SQI value SQI based on the respiratory quality index RQI RQI :

[0039]

[0040] The sensors include millimeter wave FMCW radar and WiFi wireless network card, SQI represents signal quality index, maxRQI represents the highest RQI value of a sensor among all sensors, the 2th max RQI represents the second highest RQI value of a sensor among all sensors, the 2th min RQI represents the second lowest RQI value of a sensor among all sensors, and min RQI represents the lowest RQI value of a sensor among all sensors;

[0041] Step 4-2-2, calculate the sensor SQI value SQI based on the respiratory rate RR :

[0042]

[0043] Where RR represents respiratory rate;

[0044] Step 4-2-3, calculate the sensor SQI value SQI based on the first-order difference standard deviation (SDFD) SDFD :

[0045]

[0046] Among them, minσ means that the σ of a certain sensor is the smallest, the 2th minσ means that the σ of a certain sensor is the second smallest, maxσ means that the σ of a certain sensor is the largest, and the 2th max σ means that the σ of a certain sensor is the second largest;

[0047] σ is the standard deviation of the first-order difference of the original signal:

[0048]

[0049] D[i]=x[i]-x[i-1]

[0050] Where x[i] and x[i-1] are the i-th and i-1-th sampling points of the original respiratory signal, D[i] is the difference between adjacent sampling points, N is the length of the first-order difference of the signal, and μ is the mean of the first-order difference of the original signal;

[0051] Step 4-2-4, calculate the final SQI of the sensor based on the three SQI values from steps 4-2-1 to 4-2-3:

[0052] SQI=SQI RQI ×SQI RR ×SQI SDFD

[0053] Step 4-3: Dynamically select the sensor output strategy based on the final SQI of the sensor, including single sensor direct output, multi-sensor optimal output and fusion output;

[0054] Output rules of time-varying respiratory rate TRR:

[0055]

[0056] Among them, Th is the threshold of SQI of radar and WiFi signals, SQI radar SQI for radar, SQI wifi SQI, TRR for WiFi radar is the time-varying respiratory rate of the radar, TRR wifi is the time-varying breathing rate of WiFi, TRR fusion Time-varying breathing frequency for radar and WiFi fusion.

[0057] Compared with the prior art, the present invention has the following significant advantages:

[0058] (1) The proposed radar signal processing flow, based on a signal reconstruction algorithm based on non-negative matrix factorization and a range gate selection algorithm based on a breathing quality index, is used for respiratory rate detection, eliminating the effects of random body motion and varying range gate signal quality. Within a 2 BPM (beats per minute) error range, the algorithm achieves an average respiratory detection rate exceeding 98%, an approximately 8% improvement over the traditional fixed range gate method, and enhances its ability to resist random motion interference.

[0059] (2) The signal optimized by the data correction algorithm in the WiFi signal processing process proposed in the present invention fully matches the reference sensor signal. The signal processed by the phase correction algorithm eliminates the static component bias and phase jump. The RQI-MRC-PCA fusion and selection of the optimal channel output signal processing method has a small mean absolute error in respiratory frequency. Measured data show that after processing by the above algorithm, the mean absolute error (MAE) of respiratory frequency is 0.29BPM, 0.25BPM, 0.42BPM and 0.54BPM, respectively, which is much lower than that of other comparison algorithms. The mean absolute error of respiratory frequency of RQI-MRC-PCA increases only from 0.29BPM (2 meters) to 0.54BPM (5 meters) with the increase of distance, with an increase of less than 0.25BPM, indicating that it can maintain stable performance at different distances.

[0060] (3) The proposed method of output state judgment based on SQI and multi-sensor data fusion based on an improved Bayesian model fusion algorithm can accurately select the optimal respiratory rate output at the current moment based on the characteristic parameters of signal quality. After fusion, the respiratory rate detection rate of the system reached 95.3% within an error range of 1 BPM, an improvement of more than 2.6 times that of a single radar / WiFi sensor (maximum 36.4%); the mean absolute error was reduced to 0.38 BPM, a reduction of more than 95% compared to the optimal single sensor (7.9 BPM).

[0061] The present invention is further described in detail below with reference to the accompanying drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0062] Figure 1 This is a monitoring flow chart of the dual-modal home respiratory monitoring device that integrates radar and WiFi channel status information.

[0063] Figure 2 Schematic diagram of an indoor home experiment of radar-WiFi fusion in one embodiment.

[0064] Figure 3 This is a diagram of state changes based on the SQI algorithm output in one embodiment.

[0065] Figure 4 Graph showing the time-varying respiratory rate of each sensor and fusion in one embodiment. DETAILED DESCRIPTION

[0066] In order to make the purpose, technical solutions and advantages of this application more clear, the following further describes this application in detail with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain this application and are not intended to limit this application.

[0067] It should be noted that if there are descriptions involving "first", "second", etc. in the embodiments of the present invention, the descriptions of "first", "second", etc. are only for descriptive purposes and cannot be understood as indicating or implying their relative importance or implicitly indicating the number of the indicated technical features. Therefore, the features defined as "first" and "second" may explicitly or implicitly include at least one of such features. In addition, the technical solutions between the various embodiments can be combined with each other, but they must be based on the ability of ordinary technicians in this field to implement them. When the combination of technical solutions is contradictory or cannot be implemented, it should be deemed that such a combination of technical solutions does not exist and is not within the scope of protection required by the present invention.

[0068] In one embodiment, the present invention provides a dual-modal home respiratory monitoring device that integrates radar and WiFi channel status information. The device includes a radar signal acquisition and processing module, a WiFi signal acquisition and processing module, a data communication module, and a respiratory information fusion module.

[0069] The radar signal acquisition and processing module is used to detect the human breathing signal through the millimeter wave FMCW radar, and process the I / Q orthogonal echo signal obtained by the FMCW radar to obtain the breathing signal;

[0070] The WiFi signal acquisition and processing module is used to obtain the CSI signal containing respiratory activity information through the WiFi wireless network card, calculate the CSI ratio of the receiving channel to remove time-varying noise, and then extract the respiratory signal;

[0071] The data communication module is used to build a networking system based on millimeter wave FMCW radar and WiFi through different network cable transmission modes;

[0072] The respiratory information fusion module is used to dynamically fuse the respiratory signals output by the radar signal acquisition and processing module and the WiFi signal acquisition and processing module.

[0073] The principle of the device of the present invention is as follows Figure 1 shown.

[0074] Furthermore, in one embodiment, the radar signal acquisition and processing module is configured to detect a human breathing signal using a millimeter-wave FMCW radar and process the I / Q quadrature echo signal obtained by the FMCW radar to obtain a breathing signal, specifically comprising:

[0075] Step 1-1 uses a millimeter-wave FMCW radar to collect and extract in-phase and quadrature signals, or I / Q signals, within multiple range gates. Specifically, the FMCW radar's transmit synthesizer generates a linear frequency-modulated continuous wave signal, which is reflected when it encounters a human body. The millimeter-wave FMCW radar's quadrature receiver then captures the echo signal and performs quadrature mixing with the transmitted signal. The high-frequency portion is filtered out by a low-pass filter to produce an intermediate frequency (IF) signal. The IF signal is then sampled by an ADC and converted to digital. Range information is obtained through a range FFT, and phase data within the range gate is selected and stored.

[0076] Step 1-2: preprocessing the in-phase and quadrature signals, i.e., I / Q signals, within the multiple range gates to eliminate DC bias in the signals, demodulate target information, and filter out interference noise;

[0077] Step 1-3, reconstructing the multiple range gate signals pre-processed in step 1-2 using a method based on non-negative matrix factorization to eliminate random body motion interference;

[0078] Step 1-4: For the signals of the multiple range gates obtained after reconstruction in step 1-3, a range gate is selected based on the breathing quality index to output a breathing signal.

[0079] Furthermore, in some embodiments, the pre-processing of the in-phase and quadrature signals, i.e., the I / Q signals, within the multiple range gates in step 1-2 specifically includes:

[0080] Step 1-2-1, perform DC offset elimination on the I / Q signals within multiple range gates: Use the least squares circle fitting method to solve the I / Q trajectory center offset, and subtract the offset from the original I / Q signals to eliminate the DC offset;

[0081] Here, according to complex signal theory, ideal I / Q signals form a unit circle trajectory on the complex plane centered at the origin. However, due to factors such as RF front-end non-idealities, device nonlinearities, and power supply noise, DC bias is introduced in actual systems. This DC bias causes the center of the I / Q trajectory to drift around the origin, so it is necessary to eliminate the DC bias.

[0082] Step 1-2-2, use Differentiate and Cross-Multiply (DACM) to demodulate the I / Q signal;

[0083] Step 1-2-3, filter the demodulated signal through a bandpass filter.

[0084] Here, preferably, a bandpass filter with a passband range of 0.1-0.5 Hz is used to filter the demodulated signal.

[0085] Furthermore, in some embodiments, step 1-3 reconstructs the multiple range gate signals pre-processed in step 1-2 using a method based on non-negative matrix decomposition to eliminate random body motion (RBM) interference.

[0086] Step 1-3-1, performing short-time Fourier transform (STFT) on the multiple range gate signals pre-processed in step 1-2 to obtain their time-frequency spectra, and then performing non-negative matrix decomposition on the time-frequency spectra;

[0087] Here, the preprocessed multiple range gate signals contain chest motion and random body motion (RBM) interference during breathing. The amplitude of the random body motion (RBM) mapped to the radar signal is usually much stronger than the breathing motion, resulting in the required breathing signal being overwhelmed by the random body motion (RBM), making it impossible to extract breathing frequency and breathing amplitude information. The present invention uses a non-negative matrix decomposition technology to reconstruct the signal to eliminate the random body motion (RBM) interference.

[0088] Step 1-3-1 specifically includes:

[0089] Take a range gate signal x(t) from the preprocessed multiple range gate signals and perform a short-time Fourier transform (STFT) to obtain its time spectrum |X|. Where the length of the x(t) signal is N, and the parameters of the STFT are: window length L, sliding window step size S. The calculation formula for the number of time frames T is T = [] is rounded, so the dimension of |X| is L×T;

[0090] Perform non-negative matrix factorization on |X|:

[0091]

[0092] Where K is the preset number of bases, indicating that the signal is decomposed into a combination of K frequency-time bases; the dimension of the matrix W is L×K, and each column w i is a frequency basis, representing different frequency components; the dimension of matrix H is K×T, and each row is the corresponding frequency base w i The temporal intensity change of reflects the contribution of the frequency component in different time periods; i∈{1,2,…,K} represents the index of the decomposed basis;

[0093] Step 1-3-2: detect random body motion interference (RBM) based on the two characteristics of high amplitude and sparsity, and perform signal reconstruction to eliminate random body motion interference;

[0094] Here, random body movement (RBM) is a short-term high-intensity movement, and its energy in the time-frequency domain is significantly higher than that of the periodic breathing signal, resulting in the corresponding basis component h i The energy surge of random body motion (RBM) is usually dominated by a single frequency component (such as high-frequency shock), while the respiratory signal is distributed in the low-frequency band. When RBM occurs, the energy of other basis components is close to the noise level. Therefore, RBM interference is detected based on the two characteristics of ultra-high amplitude and sparsity.

[0095] Steps 1-3-2 specifically include:

[0096] Step 1-3-2-1, detect abnormally high amplitude time window;

[0097] The short-time Fourier transform divides the long-time continuous signal into multiple local time periods. For each time window: for all time basis components h in the current time window i , calculate its average energy, and use the average energy value as the first threshold. If a time base component h in the current time window i If the energy of exceeds the first threshold, the time window is marked as an abnormally high amplitude candidate; i∈{1,2,…,K};

[0098] Step 1-3-2-2, detect sparsity;

[0099] For each abnormally high amplitude candidate time window, check all other basis components h in the time window. j Is the energy of continuously lower than the preset second threshold (a minimum value)? If so, it indicates that it has sparsity, there is RBM in the time window, and the interference source is the basis component h i ;j≠i;

[0100] Step 1-3-2-3, signal time-frequency spectrum reconstruction;

[0101] Reconstruct the signal time-frequency spectrum |X| as

[0102]

[0103] Where s i Is a selection function, if the existence of RBM is detected in the i-th time window, s i =0, otherwise s i =1;

[0104] Here, by removing the time base component h that is judged to be RBM interference in a certain time window i (ie set s i = 0), by retaining the undisturbed basis components (i.e. setting s i =1), and this is done for all time windows.

[0105] In steps 1-3-2-4, the reconstructed time-frequency spectrum is combined with the original phase, and the respiratory time domain signal with random body motion interference eliminated is generated through inverse short-time Fourier transform.

[0106] The above reconstruction method is also used for the signals of other range gates to eliminate random body motion (RBM) interference.

[0107] Furthermore, in some embodiments, step 1-4, for the signals of the multiple range gates obtained after reconstruction in step 1-3, selects a range gate based on the breathing quality index to output a breathing signal, specifically including:

[0108] Step 1-4-1, signal segmentation cutting: cutting the reconstructed multiple range gate signals (each range gate corresponds to a breathing signal at a different detection distance) into continuous time segments according to a preset time length; preferably, the preset time is 30 seconds;

[0109] Step 1-4-2, calculate the respiratory quality index RQI of each time segment of each range gate respectively;

[0110] The definition of Respiratory Quality Indices (RQI) is:

[0111]

[0112] Among them, P max It is the sum of the peak value of the spectrum and the two consecutive Fourier coefficients before and after it (a total of 3 coefficients) within the respiratory frequency range (0.1-0.7 Hz). RB is the sum of all Fourier coefficients over the entire respiratory frequency range;

[0113] Here, if the signal is a sinusoidal signal, all signals are concentrated in a single spectrum, and the RQI is 1. If the signal quality is too poor, the energy will be spread across all frequencies, and the RQI will be lower. Therefore, the higher the RQI, the more concentrated the energy of the respiratory component, and the better the quality of the respiratory signal.

[0114] Step 1-4-3, dynamic selection and signal splicing: For the same time segment, calculate the respiratory quality index RQI of all range gates respectively, compare the RQI values of all range gates in the time segment, and select the range gate signal with the highest RQI as the best range gate signal for the time segment; then, splice the best range gate signals selected for each time segment in chronological order to form a continuous output to obtain the respiratory signal.

[0115] Furthermore, in one embodiment, the WiFi signal acquisition and processing module is used to obtain a CSI signal containing respiratory activity information through a WiFi wireless network card (such as the Intel 5300 network card series), calculate the CSI ratio of the receiving channel to remove time-varying noise, and then extract the respiratory signal, specifically including:

[0116] Step 2-1: Obtaining the original CSI signal containing respiratory activity information through a WiFi wireless network card. Specifically, the WiFi wireless network card (such as the Intel 5300 network card series) captures the CSI data packet containing respiratory activity information in a multi-transmitter and multi-receiver configuration. The CSI data packet is a complex matrix with dimensions N×M×P, where N is the number of transmitting antennas, M is the number of receiving antennas, and P is the number of subcarrier channels captured by the WiFi wireless network card.

[0117] Here, a 2-transmit 3-receive configuration is preferably adopted. Due to the limitations of the CSI Tool, only 30 of the 56 valid subcarriers in the IEEE 802.11n protocol are captured. The WiFi network card continuously monitors the channel and captures CSI data packets containing respiratory activity information. The data packet content is a complex matrix with dimensions of 2 (transmitting antenna) × 3 (receiving antenna) × 30 (subcarriers).

[0118] Step 2-2: Calculate the CSI ratio of the two receiving antennas to obtain a CSI ratio data packet, thereby removing time-varying noise.

[0119] For each CSI data packet corresponding to a timestamp: For each transmit antenna, calculate the CSI ratios (CSI Ratios) for all possible receive antenna pairs, generating W rows × 1 column of Ratio values. This is then used to form a W × Q CSIRatio matrix, where W is the total number of subcarriers for all possible receive antenna pairs, and Q is the number of CSI Ratio data packets.

[0120] Here, for example, the CSI Ratio is calculated for a single data packet: For the above 2-transmit 3-receive configuration, for each transmit antenna (TX1 and TX2), the CSI ratios of all possible receive antenna pair combinations are calculated, including the receive antenna pairs for TX1 (RX2 / RX1, RX3 / RX1, RX3 / RX2) and the receive antenna pairs for TX2 (RX2 / RX1, RX3 / RX1, RX3 / RX2).

[0121] For each data packet, the CSI ratios of all receiving antenna pairs are calculated, generating 180 rows × 1 column of ratio values. This ultimately forms a 180 × N matrix (N is the number of data packets), containing 180 subcarrier CSI ratios.

[0122] Step 2-3: Perform correction based on prior information, phase static component compensation, and phase jump compensation on the CSI Ratio data packet output in step 2-2, and output an updated W×Q1 CSI Ratio matrix, where Q1 is the number of updated CSI Ratio data packets.

[0123] Step 2-4: Divide the W×Q1 CSI Ratio matrix output from step 2-3 into w channels, each of which includes subcarriers; for each channel The subcarriers are combined based on the maximum ratio combining and principal component analysis method (RQI-MRC-PCA) of the RQI. Then, w channels are selected based on the respiratory quality index (RQI) to generate the final respiratory waveform.

[0124] Here, for example, the 180×M CSI Ratio matrix (M is the number of valid data packets) output by steps 2-3 is composed of 6 channels, each of which contains 30 subcarriers. The 30 subcarriers of each channel are subjected to RQI-MRC-PCA (maximum ratio combining-principal component analysis) subcarrier fusion based on RQI. Finally, the signals of the 6 channels are selected based on the respiratory quality index RQI to generate the final respiratory waveform.

[0125] Furthermore, in some embodiments, steps 2-3 specifically include:

[0126] Step 2-3-1: Correct the CSI Ratio of each subcarrier based on prior information, specifically including:

[0127] For the W × Q (180 × N) CSI Ratio matrix output in step 2-2, filter CSI Ratio packets based on time difference: traverse the timestamps of adjacent CSI Ratio packets and calculate the time difference. When the time difference is less than the preset threshold, mark it as a duplicate packet, remove one of the CSI Ratio packets, and output a W × Q1 CSI Ratio matrix, where Q1 is the total number of valid CSI Ratio packets remaining after removal. This is expressed as follows:

[0128]

[0129] CSI Ratio i =CSI Ratio i ×Remain i

[0130] Among them, Remain i is the i-th CSI Ratio data packet, namely CSI Ratio iThe retention factor, T i and T i-1 Represent the timestamps of the i-th CSI Ratio data packet and the i-1-th CSI Ratio data packet, T TH is the preset threshold, if T i -T i-1 If the value of is less than the preset threshold, the data packet sampling interval is considered abnormal and the i-th CSI Ratio data packet needs to be removed;

[0131] Step 2-3-2: Perform phase jump compensation for each subcarrier CSI Ratio, specifically including:

[0132] For each subcarrier's CSI ratio, calculate the phase difference between adjacent time sequences, take the average of all phase difference absolute values, and set 10 times the average value as the dynamic threshold. The phase difference value is obtained by subtracting the phase at the previous moment from the phase at the next moment.

[0133] Traverse the time series data. If the absolute value of the phase difference at a certain moment exceeds the dynamic threshold, it is determined to be a phase jump. The CSI Ratio data packet (real and imaginary parts) at that moment is multiplied by the compensation factor -1 to offset the jump; otherwise, the original value is retained.

[0134] It is expressed by the formula:

[0135] D phase [i]=Phase i -Phase i-1 i≥2

[0136] Phase TH =10×ave(abs(D phase ))

[0137]

[0138] real(CSI Ratio i )=real(CSI Ratio i )×Compensate i

[0139] imag(CSI Ratio i )=imag(CSI Ratio i )×Compensate i

[0140] The first formula: Calculate the phase of the i-th CSI Ratio packet i and the phase of the i-1th CSI Ratio packeti-1 The difference value is used to detect whether there is a sudden change between adjacent time series;

[0141] The second formula: According to the absolute value of the phase difference of all data packets of the current subcarrier abs(D phase ) calculate the average value (ave), and use 10 times the average value as the dynamic threshold (Phase TH );

[0142] The third formula calculates the compensation factor. The absolute value of the phase difference between two adjacent CSI Ratio packets is calculated. If this absolute value exceeds the dynamic threshold, a phase jump is determined and the compensation factor is set to -1. Otherwise, it remains at 1 (no jump, no compensation required).

[0143] The fourth and fifth formulas multiply the real and imaginary parts of the CSI Ratio at the transition moment by a compensation factor of -1 to achieve phase compensation.

[0144] Among them, Phase i and Phase i-1 Represent the phase of the i-th CSI Ratio data packet and the i-1-th CSI Ratio data packet, respectively, D phase [i] is the phase difference, Phase TH is the threshold, Compensate i is the compensation factor for the i-th CSI Ratio packet;

[0145] Step 2-3-3, for each subcarrier CSI Ratio, performs phase static component compensation, specifically including:

[0146] The circle fitting method based on the least squares method is adopted. Through iterative calculation, the errors of the horizontal and vertical coordinates and the radius are made less than the preset allowable error, and then the coordinates of the circle center and the radius are output to complete the phase static component compensation.

[0147] Here, first, the least square method is used to fit the circle of the CSI Ratio constellation. Let the coordinates of the circle center be (a, b) and the radius be r. Then the circle equation and the points on the circle (x i ,y i ) to the center of the circle is d i :

[0148] (xa) 2 +(yb) 2 =r 2

[0149]

[0150] Where (x i ,yi ) is the coordinate of a point on the circle, d i is the distance from the point to the center of the circle. According to the idea of least squares, only ∑(d i -r) 2 When is minimum, the estimated circle is the target circle fitted by the least squares method:

[0151] E=min[(∑(d i -r) 2 )]

[0152] Expand the above formula:

[0153]

[0154] The above formula calculates the partial derivatives of a, b, and r respectively, sets the partial derivatives to 0, and combines the iterative idea to obtain the iterative formula:

[0155]

[0156] Where n is the number of iterations. The iteration is performed until the error is less than the set threshold. The center (a, b) and radius r are output, and a and b are subtracted from the real and imaginary parts of CSIRatio, respectively, to complete the phase static component compensation.

[0157] Furthermore, in some embodiments, steps 2-4 specifically include:

[0158] Step 2-4-1: Perform subcarrier screening by channel processing, retain subcarriers with RQI higher than the channel average threshold, and remove the remaining subcarriers; specifically:

[0159] For each channel and each subcarrier, calculate its respiratory quality index RQI;

[0160] Calculate the average RQI value of all subcarriers in the current channel as the subcarrier rejection threshold. When the RQI value of a subcarrier is greater than the subcarrier rejection threshold, retain the subcarrier; otherwise, remove it. The calculation formula for the subcarrier retention factor is:

[0161]

[0162] Where, RQI i is the RQI value of the i-th subcarrier, RQI th is the subcarrier rejection threshold, is the retention factor of the i-th subcarrier;

[0163] Step 2-4-2, calculate the gain and direction factor of the subcarriers reserved for each channel;

[0164] (1) Based on maximum ratio combining (MRC), the gain of the subcarriers retained in each channel is calculated using the following formula:

[0165]

[0166] Where, gain i is the gain of the ith subcarrier, RMS i E(breath) and RMS i E(noise) are the root mean square of the energy of the breathing signal and noise signal of the i-th subcarrier respectively;

[0167] (2) Direction correction is performed on the subcarriers retained in each channel based on principal component analysis (PCA), specifically:

[0168] Perform principal component analysis on the retained subcarrier signal and extract the first principal component. If the first principal component is positive, the direction factor value of the subcarrier signal is 1, otherwise it is -1. The calculation formula is:

[0169]

[0170] In the formula, factor PCA is the directional factor of the i-th subcarrier, PCA i is the first principal component of the i-th subcarrier;

[0171] Step 2-4-3, perform weighted calculation on the phase and gain value, retention factor value, and direction factor value of the subcarrier signal, and fuse all subcarriers in a channel into a breathing waveform:

[0172]

[0173] Where, Phase initial is the original phase of the i-th subcarrier, Phase i is the phase of the weighted calculated ith subcarrier. fusion is the phase of the fusion output;

[0174] In step 2-4-4, the phase of the fusion output of step 2-4-3 is cut into segments according to the preset duration t, where t is in seconds. The respiratory quality index RQI value is recalculated for the t-second segments of each channel. For the w channel segments in the same time period, the channel segment with the highest RQI value is selected as the best channel segment. The best channel segments of each time period are connected in chronological order to generate the final respiratory waveform.

[0175] Furthermore, in one embodiment, the data communication module constructs a radar and WiFi-based networking system through different network cable transmission modes, specifically including:

[0176] Step 3-1, connect the radar to the host: Before the radar respiratory monitoring is running, pre-configure the radar device with a remote target IP address that matches the host. Then, use a network cable to connect the radar to the host;

[0177] Step 3-2, WiFi connection with the host, specifically including:

[0178] Step 3-2-1: Set up the environment on the server and client respectively, complete the software package installation, and perform Samba configuration such as shared folders;

[0179] In step 3-2-2, on the server side, store the CSI data collected by the WiFi respiratory monitoring system in a shared folder to achieve networking between the WiFi respiratory monitoring system terminal device and the PC. The WiFi respiratory monitoring system, as the terminal device, is responsible for collecting and transmitting respiratory data to the edge device.

[0180] Furthermore, in one embodiment, the breathing information fusion module is used to dynamically fuse the breathing signals output by the radar signal acquisition and processing module and the WiFi signal acquisition and processing module, specifically including:

[0181] Step 4-1, extracting the respiratory frequency that changes over time from the respiratory signals output by the radar signal acquisition and processing module and the WiFi signal acquisition and processing module, specifically includes:

[0182] (1) Wavelet transform to extract time-frequency information: Perform wavelet transform on the respiratory signal x(t) to obtain the values of the scale a and time b parameters and the time-frequency representation W f (a,b) expression:

[0183]

[0184] Among them, a is the scale parameter, which is inversely proportional to the frequency (the smaller a is, the higher the frequency is, the lower the frequency is); b is the time parameter, which corresponds to the time axis position; ψ(*) is the mother wavelet function, which is used for local frequency analysis. is the conjugate function of ψ(*); the wavelet transform decomposes the signal into components of different scales, and obtains the time-scale joint distribution or the time-frequency representation W f (a, b), reflects the energy distribution of the signal at time b and scale a;

[0185] (2) Calculate the candidate instantaneous frequency: For W f (a,b)≠0, the candidate instantaneous frequency ω(a,b) is calculated by the phase derivative:

[0186]

[0187] Among them, the partial derivative It represents the rate of change with respect to time b, and the symbol i represents the imaginary unit. Through the imaginary unit i, the above formula converts the phase change in the complex domain into the instantaneous frequency in the real domain;

[0188] (3) Synchronous compression transformation: transform the time-frequency representation W f (a, b) Redistribute along the frequency axis to generate a high-resolution time-frequency distribution T f (ω,b):

[0189]

[0190] Among them, A(b) is the effective scale set, satisfying |W f (a,b)|>∈, ∈ is the threshold used to suppress noise; δ(·) is the Dirac function, which redistributes the energy to the estimated instantaneous frequency ω. Through synchronous compression, the energy scattered on multiple scales is concentrated near the true instantaneous frequency, thereby improving the time-frequency resolution;

[0191] (4) Extract instantaneous respiratory frequency: For each time b, in the time-frequency distribution T f Find the frequency corresponding to the energy peak in (ω,b) as the instantaneous breathing frequency f r (t):

[0192] f r (t) = argmax(T f (ω,b))

[0193] Among them, argmax(T f (ω,b)) represents the independent variable ω corresponding to the maximum value of energy intensity;

[0194] Step 4-2: For the respiratory signals output by the radar signal acquisition and processing module and the WiFi signal acquisition and processing module, the signal quality of the sensors is comprehensively evaluated using three independent indicators: respiratory quality index, respiratory frequency, and signal fluctuation standard deviation. Specifically, the following indicators are used:

[0195] Step 4-2-1, calculate the sensor SQI value SQI based on the respiratory quality index RQI RQI :

[0196]

[0197] The sensors include millimeter wave FMCW radar and WiFi wireless network card, SQI represents signal quality index, maxRQI represents the highest RQI value of a sensor among all sensors, the 2th max RQI represents the second highest RQI value of a sensor among all sensors, the 2th min RQI represents the second lowest RQI value of a sensor among all sensors, and min RQI represents the lowest RQI value of a sensor among all sensors;

[0198] The calculation process of RQI can refer to steps 1-4-2. The above formula shows that: within the same time period, if the RQI value of a sensor ranks highest among all sensors, its SQI RQI =1 (best quality). If it ranks second highest, then SQI RQI =0.6, and so on.

[0199] Step 4-2-2, calculate the sensor SQI value SQI based on the respiratory rate RR :

[0200]

[0201] Where RR represents respiratory rate;

[0202] Here, the main frequency within a fixed time window is extracted by Fourier transform and converted into breaths / minute to calculate the respiratory rate (RR). When RR is in the range of 9–25 breaths / minute, SQI RR =1 (ideal range), set gradient values for other settings outside this range;

[0203] Step 4-2-3, calculate the sensor SQI value SQI based on the standard deviation of first-order difference (SDFD) SDFD :

[0204]

[0205] Among them, minσ means that the σ of a certain sensor is the smallest, the 2th minσ means that the σ of a certain sensor is the second smallest, maxσ means that the σ of a certain sensor is the largest, and the 2th max σ means that the σ of a certain sensor is the second largest;

[0206] Here, for the sensor with the smallest σ, it means that the sensor has the best stability, SQI SDFD Assign a value of 1, for the sensor with the second smallest σ, SQI SDFD Assign a value of 0.9, and so on;

[0207] σ is the standard deviation of the first-order difference of the original signal:

[0208]

[0209] D[i]=x[i]-x[i-1]

[0210] Where x[i] and x[i-1] are the i-th and i-1-th sampling points of the original respiratory signal, D[i] is the difference between adjacent sampling points, N is the length of the first-order difference of the signal, and μ is the mean of the first-order difference of the original signal;

[0211] Step 4-2-4, calculate the final SQI of the sensor based on the three SQI values from steps 4-2-1 to 4-2-3:

[0212] SQI=SQI RQI ×SQI RR ×SQI SDFD

[0213] Step 4-3, based on the final SQI of the sensor, dynamically select the sensor output strategy, including single sensor direct output, multi-sensor optimal output and fusion output. The algorithm effect is as follows: Figure 4 As shown;

[0214] Output rules of time-varying respiratory rate TRR:

[0215]

[0216] Among them, Th is the threshold of SQI of radar and WiFi signals, SQI radar SQI for radar, SQI wifi SQI, TRR for WiFi radar is the time-varying respiratory rate of the radar, TRR wifi is the time-varying breathing rate of WiFi, TRR fusion Time-varying breathing frequency for radar and WiFi fusion.

[0217] Here, the output rule of time-varying respiratory rate (TRR) is described:

[0218] If the dual sensors are reliable (the radar and WiFi signal SQI are both above the threshold), the fusion result TRR is output fusion ;

[0219] Only radar is reliable (radar signal SQI is above the threshold, WiFi signal SQI is below the threshold) or single radar position, directly output radar TRR radar ;

[0220] Only WiFi is reliable (WiFi signal SQI is above the threshold, radar signal SQI is below the threshold) or a single WiFi location, directly output WiFi TRR wifi ;

[0221] The dual sensors are unreliable (the SQI of both radar and WiFi signals are below the threshold), but the fusion result TRR is still output. fusion , because fusion may alleviate the problem of single sensor failure.

[0222] Here, in some embodiments, the time-varying respiratory frequency fused by the radar and WiFi is fused using a Bayesian fusion algorithm, and the specific process includes:

[0223] Assume that the kth breathing frequency measurements from radar and WiFi are BR r (k) and BR w (k), the posterior probability density function p(BR|BR) of radar and WiFi r (k)) and p(BR|BR w (k)) is;

[0224]

[0225] Based on the first 15 seconds of time-varying respiratory rate data from a reference signal (e.g., a Huake respiratory sensor), whose probability density function follows an ideal Gaussian distribution (normal distribution), the parameters of the Gaussian distribution (mean η and standard deviation σ) are calculated using statistical methods. BR is the true respiratory rate of radar and WiFi.

[0226] The probability density function after fusion is p(BR|BR r (k)BR w (k)):

[0227]

[0228] Among them, the numerator is the joint likelihood probability of radar and WiFi at the current moment, and the denominator is the predicted probability of radar at the previous moment; p(BR|BR r (k-1))、p(BR|BR w (k-1)) are the posterior probability density functions corresponding to the k-1th respiratory frequency measurements of radar and WiFi respectively;

[0229] Select the respiratory frequency corresponding to the maximum value of the fusion probability density function as the output frequency f' r (t):

[0230] f' r (t)=arg max(p(BR|BR r (k)BR w(k)))ω=[0.1,0.7]Hz.

[0231] As a specific example, the present invention is further verified and explained in one of the embodiments.

[0232] This example uses respiratory rate (breaths per minute, BPM) as the evaluation metric. The number of sampling points where the absolute error between the radar and reference sensor's respiratory rate falls within 1 BPM, 1.5 BPM, and 2 BPM is counted, and then divided by the total number of sampling points to obtain the corresponding accuracy, or respiratory detection rate. The respiratory detection rates for six subjects using various algorithms are shown in Table 1.

[0233] Table 1 Breathing detection rate of subjects

[0234]

[0235] The average breath detection rates in Table 1 are shown in Table 2 .

[0236] Table 2 Average respiratory detection rate

[0237]

[0238] Combined with the quantitative experimental results in Tables 1 and 2, body motion segments have a significant impact on respiratory rate detection. After excluding manually calibrated RBMs, many algorithms maintain a respiratory detection rate of over 90% when the absolute error is within 2 BPM. However, when the RBM segments are not processed, the respiratory detection rate drops significantly, reaching only 66.4% when the absolute error is within 1 BPM. This is also a difficulty in current research on non-contact vital sign monitoring of the human body. In some cases (such as subject 4 in Table 1), the respiratory detection rate obtained based on the NMF signal reconstruction method is lower than the unprocessed respiratory detection rate. Analysis shows that this is due to damage to the normal band during signal reconstruction, resulting in a decrease in signal quality.

[0239] From the above, it can be seen that the signal reconstruction algorithm based on non-negative matrix decomposition and the range gate selection algorithm based on respiratory quality index proposed in the present invention are used for respiratory rate detection. Compared with the original algorithm, the accuracy is greatly improved. The average respiratory detection rate of the NMF+RQI algorithm reaches more than 98% at 2BPM.

[0240] In this embodiment, the positions and sensors corresponding to the output states of the algorithm of the present invention are shown in Table 3. Figure 2 shown.

[0241] Table 3 Output status corresponding to the position and sensor

[0242]

[0243] from Figure 3 As can be seen from Table 3, the algorithm always selects one optimal sensor to output alone or fuses the results of two optimal sensors to output.

[0244] In stage 1, the algorithm selects radar 1 results and outputs them separately, which is state 1.

[0245] In stage 2, both Radar 2 and WiFi 1 contain valid respiration data, with WiFi 1 having better signal quality. When the SQI of both Radar 2 and WiFi 1 signals exceeds the threshold, the algorithm performs Bayesian fusion on the results of both sensors, resulting in state 2 (the red circled portion between 251 and 300 seconds in the figure). Otherwise, the results from WiFi 1, which has better signal quality, are output alone, resulting in state 3.

[0246] The algorithm in stage 3 selects WiFi 2 and outputs the result separately, which is state 4.

[0247] The SQI-based algorithm output state selection algorithm can accurately select the optimal respiratory rate output at the current moment based on the characteristic parameters of signal quality.

[0248] In this embodiment, the algorithm of the present invention is used to conduct a respiratory monitoring experiment on a multi-sensor network, and the results are shown in Table 4.

[0249] Table 4 Respiratory monitoring test results

[0250]

[0251] As shown in Table 4, the detection rates within 1 BPM for Radar 1, Radar 2, WiFi 1, WiFi 2, and sensor fusion were 33.7%, 21.6%, 36.4%, 31.5%, and 95.3%, respectively. Their overall mean absolute errors were 9.9 BPM, 11.1 BPM, 7.9 BPM, 9.6 BPM, and 0.38 BPM, respectively. The sensor fusion output had the highest respiration detection rate and the lowest overall mean absolute error, demonstrating the effectiveness of the proposed algorithm in detecting time-varying respiratory rate.

[0252] In summary, the signal reconstruction algorithm based on non-negative matrix decomposition and the range gate selection algorithm based on the respiratory quality index proposed in the present invention, the data correction algorithm based on prior information, the phase correction algorithm based on iterative circle fitting, the subcarrier fusion algorithm based on RQI-MRC-PCA, the long-term channel switching algorithm based on the respiratory quality index, the output state judgment based on SQI and the multi-sensor data fusion method based on the improved Bayesian model fusion algorithm, each algorithm is effective and feasible, and the performance is reliable.

[0253] The above shows and describes the basic principles, main features and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The above embodiments and descriptions are only illustrative of the principles of the present invention. Without departing from the spirit and scope of the present invention, any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention shall be included in the scope of protection of the present invention.

Claims

1. A dual-modal home respiratory monitoring device integrating radar and WiFi channel status information, characterized in that: The device includes a radar signal acquisition and processing module, a WiFi signal acquisition and processing module, a data communication module and a respiratory information fusion module; The radar signal acquisition and processing module is used to detect the human breathing signal through the millimeter wave FMCW radar, and process the I / Q orthogonal echo signal obtained by the FMCW radar to obtain the breathing signal; The WiFi signal acquisition and processing module is used to obtain the CSI signal containing respiratory activity information through the WiFi wireless network card, calculate the CSI ratio of the receiving channel to remove time-varying noise, and then extract the respiratory signal; The data communication module is used to build a networking system based on millimeter wave FMCW radar and WiFi through different network cable transmission modes; The respiratory information fusion module is used to dynamically fuse the respiratory signals output by the radar signal acquisition and processing module and the WiFi signal acquisition and processing module.

2. The dual-modal home respiratory monitoring device integrating radar and WiFi channel status information according to claim 1 is characterized in that: The radar signal acquisition and processing module is used to detect the human breathing signal through the millimeter wave FMCW radar and process the I / Q orthogonal echo signal obtained by the FMCW radar to obtain the breathing signal, specifically including: Step 1-1 uses a millimeter-wave FMCW radar to collect and extract in-phase and quadrature signals, or I / Q signals, within multiple range gates. Specifically, the FMCW radar's transmit synthesizer generates a linear frequency-modulated continuous wave signal, which is reflected when it encounters a human body. The millimeter-wave FMCW radar's quadrature receiver then captures the echo signal and performs quadrature mixing with the transmitted signal. The high-frequency portion is filtered out by a low-pass filter to produce an intermediate frequency (IF) signal. The IF signal is then sampled by an ADC and converted to digital. Range information is obtained through a range FFT, and phase data within the range gate is selected and stored. Step 1-2: preprocessing the in-phase and quadrature signals, i.e., I / Q signals, within the multiple range gates to eliminate DC bias in the signals, demodulate target information, and filter out interference noise; Step 1-3, reconstructing the multiple range gate signals pre-processed in step 1-2 using a method based on non-negative matrix factorization to eliminate random body motion interference; Step 1-4: For the signals of the multiple range gates obtained after reconstruction in step 1-3, a range gate is selected based on the breathing quality index to output a breathing signal.

3. The dual-modal home respiratory monitoring device integrating radar and WiFi channel status information according to claim 2, characterized in that: Step 1-2 of preprocessing the in-phase and quadrature signals, i.e., the I / Q signals, within the multiple range gates specifically includes: Step 1-2-1, perform DC offset elimination on the I / Q signals within multiple range gates: Use the least squares circle fitting method to solve the I / Q trajectory center offset, and subtract the offset from the original I / Q signals to eliminate the DC offset; Step 1-2-2, demodulate the I / Q signal using the differential cross-multiplication method; Step 1-2-3, filter the demodulated signal through a bandpass filter.

4. The dual-modal home respiratory monitoring device integrating radar and WiFi channel status information according to claim 2, characterized in that: Step 1-3 reconstructs the multiple range gate signals pre-processed in step 1-2 using a method based on non-negative matrix factorization to eliminate random body motion interference, specifically including: Step 1-3-1, performing a short-time Fourier transform on the multiple range gate signals pre-processed in step 1-2 to obtain their time-frequency spectra, and then performing non-negative matrix decomposition on the time-frequency spectra; specifically, comprising: Take a certain range gate signal x(t) from the preprocessed multiple range gate signals and perform a short-time Fourier transform (STFT) to obtain its time spectrum |X|. Where the length of the x(t) signal is N, the parameters of the STFT are: window length L, sliding window step size S, then the calculation formula for the time frame number T is: [] is rounded, so the dimension of |X| is L×T; Perform non-negative matrix factorization on |X|: Where K is the preset number of bases, indicating that the signal is decomposed into a combination of K frequency-time bases; the dimension of the matrix W is L×K, and each column w i is a frequency basis, representing different frequency components; the dimension of matrix H is K×T, and each row is the corresponding frequency base w i The temporal intensity change of reflects the contribution of the frequency component in different time periods; i∈{1,2,…,K} represents the index of the decomposed basis; Step 1-3-2 detects random body motion interference (RBM) based on the two characteristics of high amplitude and sparsity, and reconstructs the signal to eliminate random body motion interference. Specifically, it includes: Step 1-3-2-1, detect abnormally high amplitude time window; The short-time Fourier transform divides the long-time continuous signal into multiple local time periods. For each time window: for all time basis components h in the current time window i , calculate its average energy, and use the average energy value as the first threshold. If a time base component h in the current time window i If the energy of exceeds the first threshold, the time window is marked as an abnormally high amplitude candidate; i∈{1,2,…,K}; Step 1-3-2-2, detect sparsity; For each abnormally high amplitude candidate time window, check all other basis components h in the time window. j Is the energy of continuously lower than the preset second threshold? If so, it indicates that it has sparsity, there is RBM in the time window, and the interference source is the basis component h i ;j≠i; Step 1-3-2-3, signal time-frequency spectrum reconstruction; Reconstruct the signal time-frequency spectrum |X| as Where s i Is a selection function, if the existence of RBM is detected in the i-th time window, s i =0, otherwise s i =1; In steps 1-3-2-4, the reconstructed time-frequency spectrum is combined with the original phase, and the respiratory time domain signal with random body motion interference eliminated is generated through inverse short-time Fourier transform.

5. The dual-modal home respiratory monitoring device integrating radar and WiFi channel status information according to claim 2, characterized in that: Step 1-4, for the signals of the multiple range gates obtained after the reconstruction in step 1-3, selects the range gate based on the breathing quality index and outputs the breathing signal, specifically including: Step 1-4-1, signal segment cutting: cutting the reconstructed multiple range gate signals into continuous time segments according to the preset time length; Step 1-4-2, calculate the respiratory quality index RQI of each time segment of each range gate respectively; The respiratory quality index is defined as: Among them, P max It is the sum of the peak value of the spectrum and the two consecutive Fourier coefficients before and after it within the respiratory frequency range, P RB is the sum of all Fourier coefficients over the entire respiratory frequency range; Step 1-4-3, dynamic selection and signal splicing: For the same time segment, calculate the respiratory quality index RQI of all range gates respectively, compare the RQI values of all range gates in the time segment, and select the range gate signal with the highest RQI as the best range gate signal for the time segment; then, splice the best range gate signals selected for each time segment in chronological order to form a continuous output to obtain the respiratory signal.

6. The dual-modal home respiratory monitoring device integrating radar and WiFi channel status information according to claim 1, characterized in that: The WiFi signal acquisition and processing module is used to obtain the CSI signal containing respiratory activity information through the WiFi wireless network card, calculate the CSI ratio of the receiving channel to remove time-varying noise, and then extract the respiratory signal. Specifically, it includes: Step 2-1: Obtaining an original CSI signal containing respiratory activity information through a WiFi wireless network card. Specifically, the WiFi wireless network card, in a multi-transmitter and multi-receiver configuration, captures a CSI data packet containing respiratory activity information. The CSI data packet is a complex matrix with dimensions N×M×P, where N is the number of transmitting antennas, M is the number of receiving antennas, and P is the number of subcarrier channels captured by the WiFi wireless network card. Step 2-2: Calculate the CSI ratio of the two receiving antennas to obtain a CSI ratio data packet, thereby removing time-varying noise. For each CSI data packet corresponding to a timestamp: For each transmit antenna, calculate the CSI ratios (CSI Ratios) for all possible combinations of receive antenna pairs, generating W rows × 1 column of Ratio values. This is then used to form a W × Q CSI Ratio matrix, where W is the total number of subcarriers for all possible receive antenna pairs, and Q is the number of CSI Ratio data packets. Step 2-3: Perform correction based on prior information, phase static component compensation, and phase jump compensation on the CSI Ratio data packet output in step 2-2, and output an updated W×Q1 CSI Ratio matrix, where Q1 is the number of updated CSI Ratio data packets. Step 2-4: Divide the W×Q1 CSI Ratio matrix output from step 2-3 into w channels, each of which includes subcarriers; for each channel The subcarriers are combined based on the maximum ratio combining and principal component analysis method (RQI-MRC-PCA) of the RQI. Then, w channels are selected based on the respiratory quality index (RQI) to generate the final respiratory waveform.

7. The dual-modal home respiratory monitoring device integrating radar and WiFi channel status information according to claim 6, characterized in that: Steps 2-3 specifically include: Step 2-3-1: Correct the CSI Ratio of each subcarrier based on prior information, specifically including: For the W × Q CSI Ratio matrix output in step 2-2, filter CSI Ratio packets based on time difference: traverse the timestamps of adjacent CSI Ratio packets and calculate the time difference. When the time difference is less than the preset threshold, mark it as a duplicate packet and remove one of the CSI Ratio packets. Output the W × Q1 CSI Ratio matrix, where Q1 is the total number of valid CSI Ratio packets remaining after removal. Step 2-3-2: Perform phase jump compensation for each subcarrier CSI Ratio, specifically including: For each subcarrier's CSI ratio, calculate the phase difference between adjacent time sequences, take the average of all phase difference absolute values, and set 10 times the average value as the dynamic threshold. The phase difference value is obtained by subtracting the phase at the previous moment from the phase at the next moment. Traverse the time series data. If the absolute value of the phase difference at a certain moment exceeds the dynamic threshold, it is determined to be a phase jump. The CSI Ratio data packet at that moment is multiplied by the compensation factor -1 to offset the jump; otherwise, the original value is retained. Step 2-3-3, for each subcarrier CSI Ratio, performs phase static component compensation, specifically including: The circle fitting method based on the least squares method is adopted. Through iterative calculation, the errors of the horizontal and vertical coordinates and the radius are made less than the preset allowable error, and then the coordinates of the circle center and the radius are output to complete the phase static component compensation.

8. The dual-modal home respiratory monitoring device integrating radar and WiFi channel status information according to claim 5 or 6, characterized in that: Steps 2-4 specifically include: Step 2-4-1: Perform subcarrier screening by channel processing, retain subcarriers with RQI higher than the channel average threshold, and remove the remaining subcarriers; specifically: For each channel and each subcarrier, calculate its respiratory quality index RQI; Calculate the average RQI value of all subcarriers in the current channel as the subcarrier rejection threshold. When the RQI value of a subcarrier is greater than the subcarrier rejection threshold, retain the subcarrier; otherwise, remove it. The calculation formula for the subcarrier retention factor is: Where, RQI i is the RQI value of the i-th subcarrier, RQI th is the subcarrier rejection threshold, is the retention factor of the i-th subcarrier; Step 2-4-2, calculate the gain and direction factor of the subcarriers reserved for each channel; (1) Based on maximum ratio combining (MRC), the gain of the subcarriers retained in each channel is calculated using the following formula: Where, gain i is the gain of the ith subcarrier, RMS iE(breath) and RMS iE(noise) are the root mean square of the energy of the breathing signal and noise signal of the i-th subcarrier respectively; (2) Direction correction is performed on the subcarriers retained in each channel based on principal component analysis (PCA), specifically: Perform principal component analysis on the retained subcarrier signal and extract the first principal component. If the first principal component is positive, the direction factor value of the subcarrier signal is 1, otherwise it is -1. The calculation formula is: In the formula, factor PCA is the directional factor of the i-th subcarrier, PCA i is the first principal component of the i-th subcarrier; Step 2-4-3, perform weighted calculation on the phase and gain value, retention factor value, and direction factor value of the subcarrier signal, and fuse all subcarriers in a channel into a breathing waveform: Where, Phase initial is the original phase of the i-th subcarrier, Phase i is the phase of the weighted calculated i-th subcarrier, Phase fusion is the phase of the fusion output; In step 2-4-4, the phase of the fusion output of step 2-4-3 is cut into segments according to the preset duration t, where t is in seconds. The respiratory quality index RQI value is recalculated for the t-second segments of each channel. For the w channel segments in the same time period, the channel segment with the highest RQI value is selected as the best channel segment. The best channel segments of each time period are connected in chronological order to generate the final respiratory waveform.

9. The dual-modal home respiratory monitoring device integrating radar and WiFi channel status information according to claim 1, characterized in that: The respiratory information fusion module is used to dynamically fuse the respiratory signals output by the radar signal acquisition and processing module and the WiFi signal acquisition and processing module, specifically including: Step 4-1, extracting the respiratory frequency that changes over time from the respiratory signals output by the radar signal acquisition and processing module and the WiFi signal acquisition and processing module, specifically includes: (1) Wavelet transform to extract time-frequency information: Perform wavelet transform on the respiratory signal x(t) to obtain the values of the scale a and time b parameters and the time-frequency representation W f (a,b) expression: Among them, a is the scale parameter, which is inversely proportional to the frequency; b is the time parameter, corresponding to the time axis position; ψ(*) is the mother wavelet function, which is used for local frequency analysis. is the conjugate function of ψ(*); the wavelet transform decomposes the signal into components of different scales, and obtains the time-scale joint distribution or the time-frequency representation W f (a, b), reflects the energy distribution of the signal at time b and scale a; (2) Calculate the candidate instantaneous frequency: For W f (a,b)≠0, the candidate instantaneous frequency ω(a,b) is calculated by the phase derivative: Among them, the partial derivative It represents the rate of change with respect to time b, and the symbol i represents the imaginary unit; (3) Synchronous compression transformation: transform the time-frequency representation W f (a, b) Redistribute along the frequency axis to generate a high-resolution time-frequency distribution T f (ω,b): Among them, A(b) is the effective scale set, satisfying |W f (a,b)|>∈, ∈ is the threshold used to suppress noise; δ(·) is the Dirac function that redistributes energy to the estimated instantaneous frequency ω; (4) Extract instantaneous respiratory frequency: For each time b, in the time-frequency distribution T f Find the frequency corresponding to the energy peak in (ω,b) as the instantaneous breathing frequency f r (t): f r (t)=arg max(T f (ω,b)) Among them, arg max(T f (ω,b)) represents the independent variable ω corresponding to the maximum value of energy intensity; Step 4-2: For the respiratory signals output by the radar signal acquisition and processing module and the WiFi signal acquisition and processing module, the signal quality of the sensors is comprehensively evaluated using three independent indicators: respiratory quality index, respiratory frequency, and signal fluctuation standard deviation. Specifically, the following indicators are used: Step 4-2-1, calculate the sensor SQI value SQI based on the respiratory quality index RQI RQI : The sensors include millimeter wave FMCW radar and WiFi wireless network card, SQI represents signal quality index, max ROI represents the highest RQI value of a sensor among all sensors, the 2th max RQI represents the second highest RQI value of a sensor among all sensors, the 2th min RQI represents the second lowest RQI value of a sensor among all sensors, and min RQI represents the lowest RQI value of a sensor among all sensors; Step 4-2-2, calculate the sensor SQI value SQI based on the respiratory rate RR : Where RR represents respiratory rate; Step 4-2-3, calculate the sensor SQI value SQI based on the first-order difference standard deviation (SDFD) SDFD : Among them, minσ means that the σ of a certain sensor is the smallest, the 2th min σ means that the σ of a certain sensor is the second smallest, maxσ means that the σ of a certain sensor is the largest, and the 2th max σ means that the σ of a certain sensor is the second largest; σ is the standard deviation of the first-order difference of the original signal: D[i]=x[i]-x[i-1] Where x[i] and x[i-1] are the i-th and i-1-th sampling points of the original respiratory signal, D[i] is the difference between adjacent sampling points, N is the length of the first-order difference of the signal, and μ is the mean of the first-order difference of the original signal; Step 4-2-4, calculate the final SQI of the sensor based on the three SQI values from steps 4-2-1 to 4-2-3: SQI=SQI RQI ×SQI RR ×SQI SDFD Step 4-3: Dynamically select the sensor output strategy based on the final sensor signal quality index SQI, including single sensor direct output, multi-sensor optimal output and fusion output; Output rules of time-varying respiratory rate TRR: Among them, Th is the threshold of SQI of radar and WiFi signals, SQI radar SQI for radar, SQI wifi SQI, TRR for WiFi radar is the time-varying respiratory rate of the radar, TRR wifi is the time-varying respiratory rate of WiFi, TRR fusion Time-varying breathing frequency for radar and WiFi fusion.

10. The dual-modal home respiratory monitoring device integrating radar and WiFi channel status information according to claim 9, characterized in that: The time-varying respiratory frequency of the radar and WiFi fusion is fused through a Bayesian fusion algorithm. The specific process includes: Assume that the kth breathing frequency measurements from radar and WiFi are BR r (k) and BR w (k), the posterior probability density function p(BR|BR) of radar and WiFi r (k)) and p(BR|BR w (k)) is; The posterior probability density function obeys an ideal Gaussian distribution. Using the corresponding data fragments, the parameters of the Gaussian distribution, including the mean η and standard deviation σ, are calculated by statistical methods. The probability density function after fusion is p(BR|BR r (k)BR w (k)): Among them, the numerator is the joint likelihood probability of radar and WiFi at the current moment, and the denominator is the predicted probability of radar at the previous moment; p(BR|BR r (k-1))、p(BR|BR w (k-1)) are the posterior probability density functions corresponding to the k-1th respiratory frequency measurements of radar and WiFi respectively; BR is the true respiratory frequency of radar and WiFi; Select the respiratory frequency corresponding to the maximum value of the fusion probability density function as the output frequency f' r (t): f' r (t)=arg max(p(BR|BR r (k)BR w (k)))ω=[0.1,0.7]Hz。

Images