A sweat amount monitoring system and method based on RFID passive multi-frequency sensing

Abstract

The application discloses a kind of sweat quantity monitoring system and method based on RFID passive multi-frequency sensing, comprising: passive sweat quantity sensor, for receiving the wireless radio frequency signal transmitted by sweat quantity sensing integrated machine, and after modulating the received wireless radio frequency signal, send to sweat quantity sensing integrated machine;Sweat quantity sensing integrated machine is used to generate multi-frequency wireless radio frequency signal to activate passive sweat quantity sensor, and real-time acquisition wireless radio frequency signal scattered back by passive sweat quantity sensor, which is sent to data processing platform;Data processing platform, real-time acquisition phase data and RSSI characteristic data sent by sweat quantity sensing integrated machine, utilize the method of multi-frequency signal difference, extract the feature related to sweat quantity, carry out sweat quantity prediction, obtain sweat quantity.The application adopts RFID passive sensing, uses cheap RFID label, and does not need battery power supply, realizes accurate sweat quantity monitoring by analyzing the change of RFID signal.

Description

Technical Field

[0001] This invention belongs to the field of RFID passive sensing technology, specifically relating to a sweat volume monitoring system and method based on RFID passive multi-frequency sensing. Background Technology

[0002] With the development of health monitoring technology, real-time monitoring systems based on biosignals are increasingly widely used in personal health management and medical diagnosis. Human perspiration, as an important physiological indicator, reflects an individual's metabolic level, fluid balance, and potential health problems, especially in high-temperature environments, during exercise, and in monitoring specific diseases. However, current research and technological applications related to sweat mainly focus on the detection of sweat components, such as the analysis of electrolytes, glucose, lactic acid, and pH levels. While this information can reveal specific metabolic states, it often overlooks perspiration volume—a fundamental yet crucial health indicator.

[0003] Current research and technologies for detecting sweat volume are relatively limited. Even though a few studies have attempted to develop solutions for monitoring sweat volume, the following significant shortcomings remain:

[0004] (1) Reliance on carefully designed dedicated sensors: usually requires complex manufacturing processes and precise calibration, resulting in high equipment costs and unsuitability for large-scale applications.

[0005] (2) Energy limitation: Most of the existing work relies on active equipment, which requires regular charging or battery replacement. Especially during long-term use or high-frequency monitoring, the power supply problem becomes the main bottleneck.

[0006] Therefore, based on the above considerations, there is an urgent need for a low-cost, simple, and passive technology that can overcome the limitations of traditional methods and provide a new means of monitoring sweating for daily health management. Summary of the Invention

[0007] To address the shortcomings of the existing technologies, the present invention aims to provide a sweat volume monitoring system and method based on RFID passive multi-frequency sensing, thereby solving the problems of short battery life and high cost associated with existing sweat volume detection technologies that rely on active dedicated sensors. The present invention employs RFID passive sensing, uses inexpensive RFID tags, and requires no battery power, achieving accurate sweat volume monitoring by analyzing changes in RFID signals.

[0008] To achieve the above objectives, the technical solution adopted by the present invention is as follows:

[0009] This invention discloses a sweat volume monitoring system based on RFID passive multi-frequency sensing, comprising: a passive sweat volume sensor, an integrated sweat volume sensing unit, and a data processing platform; wherein,

[0010] The passive sweat volume sensor is used to receive the wireless radio frequency signal emitted by the sweat volume sensing all-in-one machine, and modulate the received wireless radio frequency signal before sending it to the sweat volume sensing all-in-one machine.

[0011] The sweat volume sensing all-in-one machine is used to generate multi-frequency wireless radio frequency signals to activate the passive sweat volume sensor, and to collect the wireless radio frequency signals sent by the passive sweat volume sensor in real time and send them to the data processing platform.

[0012] The data processing platform acquires phase data and RSSI feature data from the wireless radio frequency signal sent by the sweat volume sensing all-in-one machine in real time. It uses the multi-frequency signal differential method to extract sweat volume-related features, predicts sweat volume, and obtains the sweat volume.

[0013] Furthermore, the passive sweat volume sensor includes: an RFID tag pair and a sponge block for storing sweat. The RFID tag pair includes a sensing tag and a reference tag, wherein the sensing tag is used to sense the amount of sweat, and the reference tag is placed 1 cm away from the sensing tag to eliminate dynamic interference. The sensing tag and the reference tag are both attached to the front of the sponge block with double-sided tape, and the back of the sponge block is attached to the sweat monitoring target location.

[0014] Furthermore, the sweat volume sensing integrated machine includes: an RFID antenna and an RFID reader; the RFID reader continuously transmits multi-frequency wireless radio frequency signals through the RFID antenna, and collects the wireless radio frequency signals backscattered by the passive sweat volume sensor in real time, and sends them to the data processing platform.

[0015] Furthermore, the RFID reader and RFID tag are of UHF specification, with a frequency of 860-960MHz, and the RFID reader reads and writes the RFID tag through the EPC Global C1 G2 protocol.

[0016] Furthermore, the multi-frequency radio frequency signal is an OFDM symbol, and the specific generation steps are as follows:

[0017] Generate a random sequence S of length n, where each element has a 50% probability of being +1 or -1, and n is a user-defined OFDM symbol length;

[0018] Performing an inverse Fourier transform (IFFT) on a random sequence S yields an OFDM symbol x in the time domain. The OFDM symbol consists of n complex numbers, denoted as x = {c1, c2, ..., cn}. i ,…,c n}, where each complex number c i Represents a specific frequency f i Signal;

[0019] A cyclic suffix CS of length Q is appended to the end of the generated OFDM symbol x, where CS is the first Q elements of x, and CS = {c1, c2, ..., c...}. i ,…,c Q};

[0020] The OFDM symbol after adding CS is multiplied by an amplification factor α. The amplitude of the OFDM symbol is proportional to its energy, ensuring that the OFDM symbol has enough energy to activate the RFID tag. The value of α is 6.

[0021] Furthermore, during the EPC phase of RFID signal transmission, the sweat volume sensing device replaces the single-frequency continuous wave (CW) with an OFDM symbol containing multi-frequency information to activate the passive sweat volume sensor.

[0022] Furthermore, the method by which the data processing platform calculates the sweat volume result data is as follows:

[0023] (1) Acquire phase data of the reference tag and the sensing tag at different radio frequency signal frequencies, denoted as follows: and in For reference labels at frequency f i The phase data below, To perceive the tag at frequency f i The phase data below;

[0024] (2) Calculate the phase data W of the sensing tag. s Phase data W with reference label r The difference W d , recorded as in For frequency f i The phase data difference between the lower sensing label and the reference label;

[0025] (3) Calculate the difference W d adjacent frequencies f i with f i+1 The phase data difference between them is represented as the sweat volume characteristic X.

[0026] (4) Calculate the characteristics of all sweat volume within the measurement range;

[0027] (5) Use all the sweat volume features from step (4) above to train the regressor;

[0028] (6) Use the regressor obtained after training to predict sweat volume.

[0029] Furthermore, step (1) specifically includes;

[0030] (11) Based on the wireless radio frequency signal amplitude threshold A thresh and duration threshold T thresh Separate the EPC stage signal from the received radio frequency signal, where A thresh T is 0.05. thresh It is 3;

[0031] (12) Subcarrier alignment is performed on the OFDM symbol x in the EPC stage signal, specifically including:

[0032] (121) Perform a Fast Fourier Transform (FFT) on x to obtain the frequency domain representation. as follows:

[0033]

[0034] Where E is the subcarrier offset of the OFDM symbol, X n The random sequence used to generate OFDM symbols, where j is the imaginary unit and N is the length of the OFDM symbol;

[0035] (122) Compare with the initial X n and The phase offset is used to obtain the subcarrier offset E;

[0036] (123) Shift x by E units to obtain the OFDM symbol after subcarrier offset compensation and alignment, denoted as x. align ;

[0037] (13) x align Amplitude A x The amplitude A of the OFDM symbol during the T1 phase of RFID signal transmission T Compare them, if A x =A T Then x align When it is in the OFF state, let it be denoted as x. off Otherwise, it is in the ON state, denoted as x. on ;

[0038] (14) Based on the obtained x off and x on The phase data w is extracted using the following expression:

[0039]

[0040] Where arg represents the angle used to calculate the complex number.

[0041] Furthermore, step (2) specifically includes:

[0042] The phase data of the sensing tag is set as: (t) s1 W s1 ),(t s2 W s2 ),…,(t sm W sm ), where t s1 ,t s2 ,…,t sm W represents the time point of perception labeling. s1 W s2 ,…,W sm This indicates the corresponding phase value; the phase data for the reference label is: (t) r1 W r1 ),(t r2 W r2 ),…,(t rm W rm ), where t r1 ,t r2 ,…,t rm W represents the time point of reference label. r1 W r2 ,…,W rm Indicates the corresponding phase value; two time series t r and t s To address the inconsistency, the phase data is interpolated to map the phase data of the sensing tag to the reference tag's time point. Linear interpolation is then used to estimate the phase value of the sensing tag at the reference tag's time point, as follows:

[0043] (21) Determine the time point of the reference label: The time point of the reference label is t r1 ,t r2 ,…,t rm , indicating the desired alignment point in time;

[0044] (22) Time interval for finding sensory tags: for each reference time point t rk Find the time series t of the sensory label s Two time points that satisfy the following relationship:

[0045] t sj ≤t rk ≤t sj+1 j∈[1,m-1]

[0046] Among them, t sj and t sj+1 For t rk The two closest points in time;

[0047] (23) According to the definition of linear interpolation, the sensor tag phase value Ws (t rk At time point t rk The estimated values ​​at this location are as follows:

[0048]

[0049] Among them, W sj W sj+1 They are time points t sj and t sj+1 The phase value of the sensory tag at the location, t sj t sj+1 They are time points t rk The most recent time points;

[0050] (24) Iterate through all reference time points: For each reference label time point t rk In the time series of the sensory labels, the two most recent time points t were found. sj and t sj+1 The corresponding sensory tag phase value W is calculated using a linear interpolation formula. s (t rk );

[0051] (25) After aligning the time points of the perception label and the reference label through interpolation, for each time point t i Calculate W d (t i ) = W s (t i )-W r (t i ), to put all W d (t i The average of the values ​​is used to obtain the difference W. d W s (t i ) is the perception label at time point t i The phase value, W r (t i () serves as the reference label at time point t i The phase value, W d (t i (This refers to the perceived label and the reference label at time point t) i The difference in phase values.

[0052] Furthermore, step (5) specifically includes:

[0053] (51) Standardize the sweat volume characteristic X and the corresponding sweat volume Y to zero mean and unit variance, respectively, and denot them as X std and Y std ,as follows:

[0054]

[0055] in, δX and δX are the mean and standard deviation of X, respectively. δY and δY are the mean and standard deviation of Y, respectively, and X std Y represents the standardized characteristics of sweat volume. std This is the standardized amount of sweat.

[0056] (52) Use cross-validation to compare the model performance of different fractions p and select the p with the smallest mean square error;

[0057] (53) Establish a PLSR model, project X and Y into a low-dimensional space, and maximize the correlation between them. The specific steps are as follows:

[0058] (531) For each PLSR component t h The h-th latent variable is extracted and modeled to obtain the weight matrix W, score matrix T, loading matrix P, and regression coefficient matrix c, where W = [w1, w2, ..., w k ], T=[t1,t2,…,t k ], P = [p1, p2, ..., p k ], c = [c1, c2, ..., c k Specifically:

[0059] (5311) Calculate the weight vector w of the input sweat volume feature X. h This makes component t h The variance is the largest, and the expression is as follows:

[0060] subject to |w|=1;

[0061] (5312) Calculate the h-th component t h (Score vector), expressed as follows:

[0062] t h =Xw h

[0063] (5313) Component t h Establish a regression relationship with the amount of sweat Y, and calculate the load vector c. h The expression is as follows:

[0064]

[0065] (5314) Calculate the residual matrices of X and Y, removing the influence of the current component, as shown in the following expression:

[0066] Y new =Yt h c h

[0067] Where, p h Let X be the loading matrix, and X be the loading matrix. new Y is the residual matrix of sweat volume characteristics after removing the influence of the current components. new The residual matrix of sweat volume after removing the influence of the current components;

[0068]

[0069] (5315) Repeat steps (5311)-(5314) until p principal components are extracted;

[0070] (532) Calculate the regression coefficient β, and connect the sweat volume characteristic X with the sweat volume Y, as shown in the following expression:

[0071] β=W(P T W) -1 c

[0072] The PLSR model at this time is:

[0073]

[0074] Where ∈ represents the error term, and thus the sweat volume characteristic X and the predicted sweat volume value are established. The mapping relationship between them.

[0075] Furthermore, step (6) specifically includes:

[0076] (61) Regarding the test data X test Standardization processing was performed to obtain standardized sweat volume characteristic test data X. test,std The expression is as follows:

[0077]

[0078] (62) Using the regression coefficient β for prediction, the prediction results are obtained. The expression is as follows:

[0079]

[0080] (63) The prediction results are denormalized to obtain the final predicted value of sweat volume. The expression is as follows:

[0081]

[0082] The present invention provides a method for monitoring sweat volume based on passive multi-frequency RFID sensing, which, based on the above system, includes the following steps:

[0083] 1) Attach the passive sweat volume sensor to the target location on the human body;

[0084] 2) The sweat volume sensing all-in-one machine continuously transmits multi-frequency wireless radio frequency signals and collects the wireless radio frequency signals backscattered by the passive sweat volume sensor in real time, and sends them to the data processing platform.

[0085] 3) The data processing platform calculates the phase information of the multi-frequency signals from the above-mentioned radio frequency signals;

[0086] 4) Eliminate the influence of human movement on the perception results by using dual RFID tags and multi-frequency signal phase difference method to obtain sweat volume characteristics;

[0087] 5) Calculate all sweat volume characteristics within the measurement range and train the sweat volume regressor;

[0088] 6) Input the sweat volume characteristic test data into the sweat volume regressor for prediction to obtain the sweat volume.

[0089] Furthermore, the multi-frequency radio frequency signal in step 2) is an OFDM symbol, and the specific generation steps are as follows:

[0090] 21) Generate a random sequence S of length n, where each element has a 50% probability of being +1 or -1, and n is a custom OFDM symbol length;

[0091] 22) Perform an inverse Fourier transform on the random sequence S to obtain the OFDM symbol x in the time domain. The OFDM symbol consists of n complex numbers, denoted as x = {c1, c2, ..., cn}. i ,…,c n}, where each complex number c i Represents a specific frequency f i Signal;

[0092] 23) Add a cyclic suffix CS of length Q to the end of the generated OFDM symbol x, where CS is the first Q elements of x, and CS = {c1, c2, ..., c...} i ,…,c Q};

[0093] 24) Multiply the OFDM symbol after adding CS by the amplification factor α. The amplitude of the OFDM symbol is proportional to the energy, ensuring that the OFDM symbol has enough energy to activate the RFID tag. The value of α is 6.

[0094] 25) During the EPC phase of RFID signal transmission, the sweat volume sensing integrated machine replaces the single-frequency continuous wave with an OFDM symbol with multi-frequency information to activate the passive sweat volume sensor.

[0095] Furthermore, the method for calculating the phase of the multi-frequency signal in step 3) is as follows:

[0096] 31) Based on the wireless radio frequency signal amplitude threshold A thresh and duration threshold T thresh Separate the EPC stage signal from the received radio frequency signal, where A thresh T is 0.05. thresh It is 3;

[0097] 32) Subcarrier alignment is performed on the OFDM symbol x in the EPC stage signal, specifically including:

[0098] 321) Perform a Fast Fourier Transform on x to obtain its frequency domain representation. as follows:

[0099]

[0100] Where E is the subcarrier offset of the OFDM symbol, X n The random sequence used to generate OFDM symbols, where j is the imaginary unit and N is the length of the OFDM symbol;

[0101] 322) Compare with the initial X n and The phase offset is used to obtain the subcarrier offset E;

[0102] 323) Shift x by E units to obtain the OFDM symbol after subcarrier offset compensation and alignment, denoted as x. align ;

[0103] 33) x align Amplitude A x The amplitude A of the OFDM symbol during the T1 phase of RFID signal transmission T Compare them, if A x =A T Then x align When it is in the OFF state, let it be denoted as x. off Otherwise, it is in the ON state, denoted as x. on ;

[0104] 34) Based on the obtained x off and x on The phase data w is extracted using the following expression:

[0105]

[0106] Where arg represents the angle used to calculate the complex number.

[0107] Furthermore, the step in step 4) of obtaining the sweat volume characteristics through dual RFID tags and multi-frequency signal phase difference is as follows:

[0108] The phase data of the sensing tag is set as: (t) s1 W s1 ),(t s2 W s2 ),…,(t sm W sm ), where t s1 ,t s2 ,…,t sm W represents the time point of perception labeling. s1 W s2 ,…,W sm This indicates the corresponding phase value; the phase data for the reference label is: (t) r1 W r1 ),(t r2 W r2 ),…,(t rm W rm ), where t r1 ,t r2 ,…,t rm W represents the time point of perception labeling. r1 W r2 ,…,W rm Indicates the corresponding phase value; two time series t r and t s To address the inconsistency, the phase data is interpolated to map the phase data of the sensing tag to the reference tag's time point. Linear interpolation is then used to estimate the phase value of the sensing tag at the reference tag's time point, as follows:

[0109] 41) Determine the reference label's time point: The reference label's time point is t r1 ,t r2 ,…,t rm , indicating the desired alignment point in time;

[0110] 42) Time interval for finding sensory tags: for each reference time point t rk Find the time series t of the sensory label s Two time points that satisfy the following relationship:

[0111] t sj ≤t rk ≤t sj+1 j∈[1,m-1]

[0112] Among them, tsj and t sj+1 For t rk The two closest points in time;

[0113] 43) According to the definition of linear interpolation, the phase value W of the sensing tag s (t rk At time point t rk The estimated values ​​at this location are as follows:

[0114]

[0115] Among them, W sj W sj+1 They are time points t sj and t sj+1 The phase value of the perceived tag at the location, t sj t sj+1 They are time points t rk The most recent time points;

[0116] 44) Loop through all reference time points: For each reference label time point t rk In the time series of the sensory labels, the two most recent time points t were found. sj and t sj+1 The corresponding sensory tag phase value W is calculated using a linear interpolation formula. s (t rk );

[0117] 45) After aligning the time points of the perceptual label and the reference label through interpolation, for each time point t i Calculate W d (t i ) = W s (t i )-W r (t i ), to put all W d (t i The average of the values ​​is used to obtain the difference W. d W s (t i ) is the perception label at time point t i The phase value, W r (t i () serves as the reference label at time point t i The phase value, W d (t i (This refers to the perceived label and the reference label at time point t) i The difference in phase values;

[0118] 46) Calculate the phase data W of the sensing tag. s Phase data W with reference labelr The difference W d , recorded as in For frequency f i The phase data difference between the lower sensing label and the reference label;

[0119] 47) Calculate the difference W d adjacent frequencies f i with f i+1 The phase data difference between them is represented as the sweat volume characteristic X.

[0120] Furthermore, the step of training the regressor in step 5) is as follows:

[0121] 51) Standardize the sweat volume characteristic X and the corresponding sweat volume Y to zero mean and unit variance, respectively, and denote them as Xmean and Xvariable. std and Y std ,as follows:

[0122]

[0123] in, δX and δX are the mean and standard deviation of X, respectively. δY and δY are the mean and standard deviation of Y, respectively, and X std Y represents the standardized characteristics of sweat volume. std This is the standardized amount of sweat.

[0124] 52) Use cross-validation to compare the model performance of different fractions p and select the p with the smallest mean squared error;

[0125] 53) Establish a PLSR model, projecting X and Y into a low-dimensional space while maximizing the correlation between them. The specific steps are as follows:

[0126] 531) For each PLSR component t h The h-th latent variable is extracted and modeled to obtain the weight matrix W, score matrix T, loading matrix P, and regression coefficient matrix c, where W = [w1, w2, ..., w k ], T=[t1,t2,…,t k ], P = [p1, p2, ..., p k ], c = [c1, c2, ..., c k Specifically:

[0127] 5311) Calculate the weight vector w of the input sweat volume feature X. h This makes component th The variance is the largest, and the expression is as follows:

[0128] subject to |w|=1;

[0129] 5312) Calculate the h-th component t h (Score vector), expressed as follows:

[0130] t h =Xw h

[0131] 5313) Component t h Establish a regression relationship with the amount of sweat Y, and calculate the load vector c. h The expression is as follows:

[0132]

[0133] 5314) Calculate the residual matrices of X and Y, removing the influence of the current component, as shown in the following expression:

[0134] Y new =Yt h c h

[0135] Where, p h Let X be the loading matrix, and X be the loading matrix. new Y is the residual matrix of sweat volume characteristics after removing the influence of the current components. new The residual matrix of sweat volume after removing the influence of the current components;

[0136]

[0137] 5315) Repeat steps 5311)-5314) until p principal components are extracted;

[0138] 532) Calculate the regression coefficient β, connecting the sweat volume characteristic X with the sweat volume Y, as shown in the following expression:

[0139] β=W(P T W) -1 c

[0140] The PLSR model at this time is:

[0141]

[0142] Where ∈ represents the error term, and thus the sweat volume characteristic X and the predicted sweat volume value are established. The mapping relationship between them.

[0143] Furthermore, step 6) specifically includes:

[0144] 61) Regarding the test data X test Standardization processing was performed to obtain standardized sweat volume characteristic test data X. test,std The expression is as follows:

[0145]

[0146] 62) Use the regression coefficient β to make predictions and obtain the prediction results. The expression is as follows:

[0147]

[0148] 63) The prediction results are destandardized to obtain the final predicted value of sweat volume. The expression is as follows:

[0149]

[0150] The beneficial effects of this invention are:

[0151] 1. Dynamic environment robustness: This invention utilizes a dual-tag and multi-frequency signal phase difference method to effectively eliminate the interference of human movement on RFID radio frequency signals.

[0152] 2. Multi-dimensional feature extraction: This invention uses OFDM technology to provide multi-frequency signals, which facilitates the extraction of multi-frequency and multi-dimensional features, increasing the accuracy and robustness of sweat volume perception.

[0153] 3. Lightweight sweat volume prediction: This invention uses the PLSR lightweight regression predictor to ensure the ease of training and real-time performance of sweat volume prediction inference.

[0154] 4. Easy to deploy and low cost: This invention uses commercial RFID tags for sensing, which can be easily deployed in any location at a low cost, making it easy to deploy in large quantities. Attached Figure Description

[0155] Figure 1 This is an architecture diagram of the system of the present invention;

[0156] Figure 2 This is a schematic diagram illustrating the principle of sweat volume sensing. Detailed Implementation

[0157] To facilitate understanding by those skilled in the art, the present invention will be further described below with reference to embodiments and accompanying drawings. The content mentioned in the embodiments is not intended to limit the present invention.

[0158] Reference Figure 1 , Figure 2As shown, the present invention discloses a sweat volume monitoring system based on RFID passive multi-frequency sensing, comprising: a passive sweat volume sensor, a sweat volume sensing integrated machine, and a data processing platform; wherein,

[0159] The passive sweat volume sensor is used to receive the wireless radio frequency signal emitted by the sweat volume sensing all-in-one machine, and modulate the received wireless radio frequency signal before sending it to the sweat volume sensing all-in-one machine.

[0160] The passive sweat volume sensor includes: an RFID tag pair and a sponge block for storing sweat. The RFID tag pair includes a sensing tag and a reference tag. The sensing tag is used to sense the amount of sweat, and the reference tag is placed 1 cm away from the sensing tag to eliminate dynamic interference. The sensing tag and the reference tag are both attached to the front of the sponge block with double-sided tape, and the back of the sponge block is attached to the sweat monitoring target location (such as underarm or back).

[0161] The sweat volume sensing all-in-one machine is used to generate multi-frequency wireless radio frequency signals to activate the passive sweat volume sensor, and to collect the wireless radio frequency signals backscattered by the passive sweat volume sensor in real time and send them to the data processing platform.

[0162] The sweat volume sensing integrated machine includes an RFID antenna and an RFID reader; the RFID reader continuously transmits multi-frequency wireless radio frequency signals through the RFID antenna and collects the wireless radio frequency signals backscattered by the passive sweat volume sensor in real time, and sends them to the data processing platform.

[0163] Specifically, the RFID reader and RFID tag are of UHF specification, with a frequency of 860-960MHz, and the RFID reader reads and writes the RFID tag through the EPC Global C1 G2 protocol.

[0164] Specifically, the multi-frequency radio frequency signal is an OFDM symbol, and the specific generation steps are as follows:

[0165] Generate a random sequence S of length n, where each element has a 50% probability of being +1 or -1, and n is a user-defined OFDM symbol length;

[0166] Performing an inverse Fourier transform (IFFT) on a random sequence S yields an OFDM symbol x in the time domain. The OFDM symbol consists of n complex numbers, denoted as x = {c1, c2, ..., cn}. i ,…,c n}, where each complex number c i Represents a specific frequency f i Signal;

[0167] A cyclic suffix CS of length Q is appended to the end of the generated OFDM symbol x, where CS is the first Q elements of x, and CS = {c1, c2, ..., c...}. i ,…,c Q};

[0168] The OFDM symbol after adding CS is multiplied by an amplification factor α. The amplitude of the OFDM symbol is proportional to its energy, ensuring that the OFDM symbol has enough energy to activate the RFID tag. The value of α is 6.

[0169] In the EPC phase of RFID signal transmission, the sweat volume sensing device replaces the single-frequency continuous wave (CW) with an OFDM symbol with multi-frequency information to activate the passive sweat volume sensor.

[0170] The data processing platform acquires phase data and RSSI feature data from the wireless radio frequency signal sent by the sweat volume sensing all-in-one machine in real time. It uses the multi-frequency signal differential method to extract sweat volume-related features, predicts sweat volume, and obtains the sweat volume.

[0171] The method used by the data processing platform to calculate the sweat volume data is as follows:

[0172] (1) Acquire phase data of the reference tag and the sensing tag at different radio frequency signal frequencies, denoted as follows: and in For reference labels at frequency f i The phase data below, To perceive the tag at frequency f i The phase data below;

[0173] (2) Calculate the phase data W of the sensing tag. s Phase data W with reference label r The difference W d , recorded as in For frequency f i The phase data difference between the lower sensing label and the reference label;

[0174] (3) Calculate the difference W d adjacent frequencies f i with f i+1 The phase data difference between them is represented as the sweat volume characteristic X.

[0175] (4) Calculate the characteristics of all sweat volume within the measurement range;

[0176] (5) Use all the sweat volume features from step (4) above to train the regressor;

[0177] (6) Use the regressor obtained after training to predict sweat volume.

[0178] Specifically, step (1) includes:

[0179] (11) Based on the wireless radio frequency signal amplitude threshold A thresh and duration threshold T thresh Separate the EPC stage signal from the received radio frequency signal, where A thresh T is 0.05. thresh It is 3;

[0180] (12) Subcarrier alignment is performed on the OFDM symbol x in the EPC stage signal, specifically including:

[0181] (121) Perform a Fast Fourier Transform (FFT) on x to obtain the frequency domain representation. as follows:

[0182]

[0183] Where E is the subcarrier offset of the OFDM symbol, X n The random sequence used to generate OFDM symbols, where j is the imaginary unit and N is the length of the OFDM symbol;

[0184] (122) Compare with the initial X n and The phase offset is used to obtain the subcarrier offset E;

[0185] (123) Shift x by E units to obtain the OFDM symbol after subcarrier offset compensation and alignment, denoted as x. align ;

[0186] (13) x align Amplitude A x The amplitude A of the OFDM symbol during the T1 phase of RFID signal transmission T Compare them, if A x =A T Then x align When it is in the OFF state, let it be denoted as x. off Otherwise, it is in the ON state, denoted as x. on ;

[0187] (14) Based on the obtained x off and x on The phase data w is extracted using the following expression:

[0188]

[0189] Where arg represents the angle used to calculate the complex number.

[0190] Specifically, step (2) includes:

[0191] The phase data of the sensing tag is set as: (t) s1 W s1 ),(t s2 W s2 ),…,(t sm W sm ), where t s1 ,t s2 ,…,t sm W represents the time point of perception labeling. s1 W s2 ,…,W sm This indicates the corresponding phase value; the phase data for the reference label is: (t) r1 W r1 ),(t r2 W r2 ),…,(t rm W rm ), where t r1 ,t r2 ,…,t rm W represents the time point of reference label. r1 W r2 ,…,W rm Indicates the corresponding phase value; two time series t r and t s To address the inconsistency, the phase data is interpolated to map the phase data of the sensing tag to the reference tag's time point. Linear interpolation is then used to estimate the phase value of the sensing tag at the reference tag's time point, as follows:

[0192] (21) Determine the time point of the reference label: The time point of the reference label is t r1 ,t r2 ,…,t rm , indicating the desired alignment point in time;

[0193] (22) Time interval for finding sensory tags: for each reference time point t rk Find the time series t of the sensory label s Two time points that satisfy the following relationship:

[0194] t sj ≤t rk ≤t sj+1 j∈[1,m-1]

[0195] Among them, tsj and t sj+1 For t rk The two closest points in time;

[0196] (23) According to the definition of linear interpolation, the sensor tag phase value W s (t rk At time point t rk The estimated values ​​at this location are as follows:

[0197]

[0198] Among them, W sj W sj+1 They are time points t sj and t sj+1 The phase value of the sensory tag at the location, t sj t sj+1 They are time points t rk The most recent time points;

[0199] (24) Iterate through all reference time points: For each reference label time point t rk In the time series of the sensory labels, the two most recent time points t were found. sj and t sj+1 The corresponding sensory tag phase value W is calculated using a linear interpolation formula. s (t rk );

[0200] (25) After aligning the time points of the perception label and the reference label through interpolation, for each time point t i Calculate W d (t i ) = W s (t i )-W r (t i ), to put all W d (t i The average of the values ​​is used to obtain the difference W. d W s (t i ) is the perception label at time point t i The phase value, W r (t i () serves as the reference label at time point t i The phase value, W d (t i (This refers to the perceived label and the reference label at time point t) i The difference in phase values.

[0201] Specifically, step (5) includes:

[0202] (51) Standardize the sweat volume characteristic X and the corresponding sweat volume Y to zero mean and unit variance, respectively, and denot them as X std and Y std ,as follows:

[0203]

[0204] in, δX and δX are the mean and standard deviation of X, respectively. δY and δY are the mean and standard deviation of Y, respectively, and X std Y represents the standardized characteristics of sweat volume. std This is the standardized amount of sweat.

[0205] (52) Use cross-validation to compare the model performance of different fractions p and select the p with the smallest mean square error;

[0206] (53) Establish a PLSR model, project X and Y into a low-dimensional space, and maximize the correlation between them. The specific steps are as follows:

[0207] (531) For each PLSR component t h The h-th latent variable is extracted and modeled to obtain the weight matrix W, score matrix T, loading matrix P, and regression coefficient matrix c, where W = [w1, w2, ..., w k ], T=[t1,t2,…,t k ], P = [p1, p2, ..., p k ], c = [c1, c2, ..., c k Specifically:

[0208] (5311) Calculate the weight vector w of the input sweat volume feature X. h This makes component t h The variance is the largest, and the expression is as follows:

[0209] subject to |w|=1;

[0210] (5312) Calculate the h-th component t h (Score vector), expressed as follows:

[0211] t h =Xw h

[0212] (5313) Component t h Establish a regression relationship with the amount of sweat Y, and calculate the load vector c. h The expression is as follows:

[0213]

[0214] (5314) Calculate the residual matrices of X and Y, removing the influence of the current component, as shown in the following expression:

[0215] Y new =Yt h c h

[0216] Where, p h Let X be the loading matrix, and X be the loading matrix. new Y is the residual matrix of sweat volume characteristics after removing the influence of the current components. new The residual matrix of sweat volume after removing the influence of the current components;

[0217]

[0218] (5315) Repeat steps (5311)-(5314) until p principal components are extracted;

[0219] (532) Calculate the regression coefficient β, and connect the sweat volume characteristic X with the sweat volume Y, as shown in the following expression:

[0220] β=W(P T W) -1 c

[0221] The PLSR model at this time is:

[0222]

[0223] Where ∈ represents the error term, and thus the sweat volume characteristic X and the predicted sweat volume value are established. The mapping relationship between them.

[0224] Specifically, step (6) includes:

[0225] (61) Regarding the test data X test Standardization processing was performed to obtain standardized sweat volume characteristic test data X. test,std The expression is as follows:

[0226]

[0227] (62) Using the regression coefficient β for prediction, the prediction results are obtained. The expression is as follows:

[0228]

[0229] (63) The prediction results are denormalized to obtain the final predicted value of sweat volume. The expression is as follows:

[0230]

[0231] The present invention provides a method for monitoring sweat volume based on passive multi-frequency RFID sensing, which, based on the above system, includes the following steps:

[0232] 1) Attach the passive sweat volume sensor to the target location on the body, such as under the armpit or on the back;

[0233] 2) The sweat volume sensing all-in-one machine continuously transmits multi-frequency wireless radio frequency signals and collects the wireless radio frequency signals backscattered by the passive sweat volume sensor in real time, and sends them to the data processing platform.

[0234] In step 2), the multi-frequency radio frequency signal is an OFDM symbol, and the specific generation steps are as follows:

[0235] 21) Generate a random sequence S of length n, where each element has a 50% probability of being +1 or -1, and n is a custom OFDM symbol length;

[0236] 22) Perform an inverse Fourier transform on the random sequence S to obtain the OFDM symbol x in the time domain. The OFDM symbol consists of n complex numbers, denoted as x = {c1, c2, ..., cn}. i ,…,c n}, where each complex number c i Represents a specific frequency f i Signal;

[0237] 23) Add a cyclic suffix CS of length Q to the end of the generated OFDM symbol x, where CS is the first Q elements of x, and CS = {c1, c2, ..., c...} i ,…,c Q};

[0238] 24) Multiply the OFDM symbol after adding CS by the amplification factor α. The amplitude of the OFDM symbol is proportional to the energy, ensuring that the OFDM symbol has enough energy to activate the RFID tag. The value of α is 6.

[0239] 25) During the EPC phase of RFID signal transmission, the sweat volume sensing integrated machine replaces the single-frequency continuous wave with an OFDM symbol with multi-frequency information to activate the passive sweat volume sensor.

[0240] 3) The data processing platform calculates the phase information of the multi-frequency signals from the collected radio frequency signals;

[0241] The method for calculating the phase of a multi-frequency signal is as follows:

[0242] 31) Based on the wireless radio frequency signal amplitude threshold Athresh and duration threshold T thresh Separate the EPC stage signal from the received radio frequency signal, where A thresh T is 0.05. thresh It is 3;

[0243] 32) Subcarrier alignment is performed on the OFDM symbol x in the EPC stage signal, specifically including:

[0244] 321) Perform a Fast Fourier Transform on x to obtain its frequency domain representation. as follows:

[0245]

[0246] Where E is the subcarrier offset of the OFDM symbol, X n The random sequence used to generate OFDM symbols, where j is the imaginary unit and N is the length of the OFDM symbol;

[0247] 322) Compare with the initial X n and The phase offset is used to obtain the subcarrier offset E;

[0248] 323) Shift x by E units to obtain the OFDM symbol after subcarrier offset compensation and alignment, denoted as x. align ;

[0249] 33) x align Amplitude A x The amplitude A of the OFDM symbol during the T1 phase of RFID signal transmission T Compare them, if A x =A T Then x align When it is in the OFF state, let it be denoted as x. off Otherwise, it is in the ON state, denoted as x. on ;

[0250] 34) Based on the obtained x off and x on The phase data w is extracted using the following expression:

[0251]

[0252] Where arg represents the angle used to calculate the complex number.

[0253] 4) Eliminate the influence of human movement on the perception results by using dual RFID tags and multi-frequency signal phase difference method to obtain sweat volume characteristics;

[0254] In step 4), the step of obtaining the sweat volume characteristic through dual RFID tags and multi-frequency signal phase difference is as follows:

[0255] The phase data of the sensing tag is set as: (t) s1 W s1 ),(t s2 W s2 ),…,(t sm W sm ), where t s1 ,t s2 ,…,t sm W represents the time point of perception labeling. s1 W s2 ,…,W sm This indicates the corresponding phase value; the phase data for the reference label is: (t) r1 W r1 ),(t r2 W r2 ),…,(t rm W rm ), where t r1 ,t r2 ,…,t rm W represents the time point of perception labeling. r1 W r2 ,…,W rm Indicates the corresponding phase value; two time series t r and t s To address the inconsistency, the phase data is interpolated to map the phase data of the sensing tag to the reference tag's time point. Linear interpolation is then used to estimate the phase value of the sensing tag at the reference tag's time point, as follows:

[0256] 41) Determine the reference label's time point: The reference label's time point is t r1 ,t r2 ,…,t rm , indicating the desired alignment point in time;

[0257] 42) Time interval for finding sensory tags: for each reference time point t rk Find the time series t of the sensory label s Two time points that satisfy the following relationship:

[0258] t sj ≤t rk ≤t sj+1 j∈[1,m-1]

[0259] Among them, t sj and t sj+1 For t rk The two closest points in time;

[0260] 43) According to the definition of linear interpolation, the phase value W of the sensing tag s (t rk At time point t rk The estimated values ​​at this location are as follows:

[0261]

[0262] Among them, W sj W sj+1 They are time points t sj and t sj+1 The phase value of the sensory tag at the location, t sj t sj+1 They are time points t rk The most recent time points;

[0263] 44) Loop through all reference time points: For each reference label time point t rk In the time series of the sensory labels, the two most recent time points t were found. sj and t sj+1 The corresponding sensory tag phase value W is calculated using a linear interpolation formula. s (t rk );

[0264] 45) After aligning the time points of the perceptual label and the reference label through interpolation, for each time point t i Calculate W d (t i ) = W s (t i )-W r (t i ), to put all W d (t i The average of the values ​​is used to obtain the difference W. d W s (t i ) is the perception label at time point t i The phase value, W r (t i () serves as the reference label at time point t i The phase value, W d (t i (This refers to the perceived label and the reference label at time point t) i The difference in phase values;

[0265] 46) Calculate the phase data W of the sensing tag. s Phase data W with reference label r The difference W d , recorded as in For frequency f i The phase data difference between the lower sensing label and the reference label;

[0266] 47) Calculate the difference W d adjacent frequencies f i with f i+1 The phase data difference between them is represented as the sweat volume characteristic X.

[0267] 5) Calculate all sweat volume characteristics within the measurement range and train the sweat volume regressor;

[0268] The step of training the regressor in step 5) is as follows:

[0269] 51) Standardize the sweat volume characteristic X and the corresponding sweat volume Y to zero mean and unit variance, respectively, and denote them as Xmean and Xvariable. std and Y std ,as follows:

[0270]

[0271] in, δX and δX are the mean and standard deviation of X, respectively. δY and δY are the mean and standard deviation of Y, respectively, and X std Y represents the standardized characteristics of sweat volume. std This is the standardized amount of sweat.

[0272] 52) Use cross-validation to compare the model performance of different fractions p and select the p with the smallest mean squared error;

[0273] 53) Establish a PLSR model, projecting X and Y into a low-dimensional space while maximizing the correlation between them. The specific steps are as follows:

[0274] 531) For each PLSR component t h The h-th latent variable is extracted and modeled to obtain the weight matrix W, score matrix T, loading matrix P, and regression coefficient matrix c, where W = [w1, w2, ..., w k ], T=[t1,t2,…,t k ], P = [p1, p2, ..., p k ], c = [c1, c2, ..., c k Specifically:

[0275] 5311) Calculate the weight vector w of the input sweat volume feature X. h This makes component th The variance is the largest, and the expression is as follows:

[0276]

[0277] 5312) Calculate the h-th component t h (Score vector), expressed as follows:

[0278] t h =Xw h

[0279] 5313) Component t h Establish a regression relationship with the amount of sweat Y, and calculate the load vector c. h The expression is as follows:

[0280]

[0281] 5314) Calculate the residual matrices of X and Y, removing the influence of the current component, as shown in the following expression:

[0282]

[0283] Where, p h Let X be the loading matrix, and X be the loading matrix. new Y is the residual matrix of sweat volume characteristics after removing the influence of the current components. new The residual matrix of sweat volume after removing the influence of the current components;

[0284]

[0285] 5315) Repeat steps 5311)-5314) until p principal components are extracted;

[0286] 532) Calculate the regression coefficient β, connecting the sweat volume characteristic X with the sweat volume Y, as shown in the following expression:

[0287] β=W(P T W) -1 c

[0288] The PLSR model at this time is:

[0289]

[0290] Where ∈ represents the error term, and thus the sweat volume characteristic X and the predicted sweat volume value are established. The mapping relationship between them.

[0291] 6) Input the sweat volume characteristic test data into the sweat volume regressor for prediction to obtain the sweat volume; specifically including:

[0292] 61) Regarding the test data Xtest Standardization processing was performed to obtain standardized sweat volume characteristic test data X. test,std The expression is as follows:

[0293]

[0294] 62) Use the regression coefficient β to make predictions and obtain the prediction results. The expression is as follows:

[0295]

[0296] 63) The prediction results are destandardized to obtain the final predicted value of sweat volume. The expression is as follows:

[0297]

[0298] This invention has many specific applications. The above description is only a preferred embodiment of this invention. It should be noted that for those skilled in the art, several improvements can be made without departing from the principle of this invention, and these improvements should also be considered within the scope of protection of this invention.

Claims

1. A sweat volume monitoring system based on RFID passive multi-frequency sensing, characterized in that, include: Passive sweat volume sensor, sweat volume sensor all-in-one machine and data processing platform; The passive sweat volume sensor is used to receive the wireless radio frequency signal emitted by the sweat volume sensing all-in-one machine, and modulate the received wireless radio frequency signal before sending it to the sweat volume sensing all-in-one machine. The sweat volume sensing all-in-one machine is used to generate multi-frequency wireless radio frequency signals to activate the passive sweat volume sensor, and to collect the wireless radio frequency signals sent by the passive sweat volume sensor in real time and send them to the data processing platform. The data processing platform acquires phase data and RSSI feature data from the wireless radio frequency signal sent by the sweat volume sensing all-in-one machine in real time, extracts sweat volume-related features, predicts sweat volume, and obtains the sweat volume. The passive sweat volume sensor includes: an RFID tag pair and a sponge block for storing sweat. The RFID tag pair includes a sensing tag and a reference tag. The sensing tag is used to sense the amount of sweat, and the reference tag is placed 1 cm away from the sensing tag to eliminate dynamic interference. The sensing tag and the reference tag are both attached to the front of the sponge block with double-sided tape, and the back of the sponge block is attached to the sweat monitoring target location. The data processing platform calculates the sweat volume data as follows: (1) Acquire phase data of the reference tag and the sensing tag at different radio frequency signal frequencies, denoted as follows: and in For reference labels at frequency f i The phase data below, To sense the label at frequency f i The phase data below; (2) Calculate the phase data W of the sensing tag. s Phase data W with reference label r The difference W d , recorded as in For frequency f i The phase data difference between the lower sensing label and the reference label; (3) Calculate the difference W d adjacent frequencies f i with f i+1 The phase data difference between them is represented as the sweat volume characteristic X. (4) Calculate the characteristics of all sweat volume within the measurement range; (5) Use all the sweat volume features from step (4) above to train the regressor; (6) Use the regressor obtained after training to predict sweat volume.

2. The sweat volume monitoring system based on RFID passive multi-frequency sensing according to claim 1, characterized in that, The multi-frequency radio frequency signal is an OFDM symbol, and the specific generation steps are as follows: Generate a random sequence S of length n, where each element has a 50% probability of being +1 or -1, and n is a user-defined OFDM symbol length; Performing an inverse Fourier transform on a random sequence S yields an OFDM symbol x in the time domain. The OFDM symbol consists of n complex numbers, denoted as x = {c1, c2, ..., cn}. i ,…,c n }, where each complex number c i Represents a specific frequency f i Signal; A cyclic suffix CS of length Q is appended to the end of the generated OFDM symbol x, where CS is the first Q elements of x, and CS = {c1, c2, ..., c...}. i ,…,c Q }; Multiplying the OFDM symbol after adding CS by the amplification factor α, the amplitude of the OFDM symbol is proportional to the energy, ensuring that the OFDM symbol has enough energy to activate the RFID tag.

3. The sweat volume monitoring system based on RFID passive multi-frequency sensing according to claim 2, characterized in that, During the EPC phase of RFID signal transmission, the sweat volume sensing integrated machine replaces the single-frequency continuous wave with an OFDM symbol with multi-frequency information to activate the passive sweat volume sensor.

4. The sweat volume monitoring system based on RFID passive multi-frequency sensing according to claim 3, characterized in that, Step (1) specifically includes: (11) Based on the wireless radio frequency signal amplitude threshold A thresh and duration threshold T thresh The EPC stage signal in the received radio frequency signal is separated. (12) Subcarrier alignment is performed on the OFDM symbol x in the EPC stage signal, specifically including: (121) Perform a Fast Fourier Transform on x to obtain the frequency domain representation. as follows: Where E is the subcarrier offset of the OFDM symbol, X n The random sequence used to generate OFDM symbols, where j is the imaginary unit and N is the length of the OFDM symbol; (122) Compare with the initial X n and The phase offset is used to obtain the subcarrier offset E; (123) Shift x by E units to obtain the OFDM symbol after subcarrier offset compensation and alignment, denoted as x. align ; (13) x align Amplitude A x The amplitude A of the OFDM symbol during the T1 phase of RFID signal transmission T Compare them, if A x =A T Then x align When it is in the OFF state, let it be denoted as x. off Otherwise, it is in the ON state, denoted as x. on ; (14) Based on the obtained x off and x on The phase data w is extracted using the following expression: Where arg represents the angle used to calculate the complex number.

5. The sweat volume monitoring system based on RFID passive multi-frequency sensing according to claim 4, characterized in that, Step (2) specifically includes: The phase data of the sensing tag is set as: (t) s1 W s1 ),(t s2 W s2 ),…,(t sm W sm ), where t s1 ,t s2 ,…,t sm W represents the time point of perception labeling. s1 W s2 ,…,W sm This indicates the corresponding phase value; the phase data for the reference label is: (t r1 W r1 ),(t r2 W r2 ),…,(t rm W rm ), where t r1 ,t r2 ,…,t rm W represents the time point of reference label. r1 W r2 ,…,W rm Indicates the corresponding phase value; two time series t r and t s To address the inconsistency, the phase data is interpolated to map the phase data of the sensing tag to the reference tag's time point. Linear interpolation is then used to estimate the phase value of the sensing tag at the reference tag's time point, as follows: (21) Determine the time point of the reference label: The time point of the reference label is t r1 ,t r2 ,…,t rm , indicating the desired alignment point in time; (22) Time interval for finding sensory tags: for each reference time point t rk Find the time series t of the sensory label s Two time points that satisfy the following relationship: t sj ≤t rk ≤t sj+1 ,j∈[1,m-1] Among them, t sj and t sj+1 For t rk The two closest points in time; (23) According to the definition of linear interpolation, the sensor tag phase value W s (t rk At time point t rk The estimated values ​​at this location are as follows: Among them, W sj W sj+1 They are time points t sj and t sj+1 The phase value of the sensory tag at the location, t sj t sj+1 They are time points t rk The most recent time points; (24) Iterate through all reference time points: For each reference label time point t rk In the time series of the sensory labels, the two most recent time points t were found. sj and t sj+1 The corresponding sensory tag phase value W is calculated using a linear interpolation formula. s (t rk ); (25) After aligning the time points of the perception label and the reference label through interpolation, for each time point t i Calculate W d (t i ) = W s (t i )-W r (t i ), to put all W d (t i The average of the values ​​is used to obtain the difference W. d W s (t i ) is the perception label at time point t i The phase value, W r (t i () serves as the reference label at time point t i The phase value, W d (t i (This refers to the perceived label and the reference label at time point t) i The difference in phase values.

6. The sweat volume monitoring system based on RFID passive multi-frequency sensing according to claim 5, characterized in that, Step (5) specifically includes: (51) Standardize the sweat volume characteristic X and the corresponding sweat volume Y to zero mean and unit variance, respectively, and denot them as X std and Y std ,as follows: in, δX and δX are the mean and standard deviation of X, respectively. δY and δY are the mean and standard deviation of Y, respectively, and X std Y represents the standardized characteristics of sweat volume. std This is the standardized amount of sweat. (52) Use cross-validation to compare the model performance of different fractions p and select the p with the smallest mean square error; (53) Establish a PLSR model, project X and Y into a low-dimensional space, and maximize the correlation between them. The specific steps are as follows: (531) For each PLSR component t h Extraction and modeling are performed to obtain the weight matrix W, score matrix T, loading matrix P, and regression coefficient matrix c, E = [w1, w2, ..., w k ], T=[t1,t2,…,t k ], P = [p1, p2, ..., p k ], c = [c1, c2, ..., c k Specifically: (5311) Calculate the weight vector w of the input sweat volume feature X. h This makes component t h The variance is the largest, and the expression is as follows: (5312) Calculate the h-th component t h The expression is as follows: t h =Xw h (5313) Component t h Establish a regression relationship with the amount of sweat Y, and calculate the load vector c. h The expression is as follows: (5314) Calculate the residual matrices of X and Y, removing the influence of the current component, as shown in the following expression: Where, p h Let X be the loading matrix, and X be the loading matrix. new Y is the residual matrix of sweat volume characteristics after removing the influence of the current components. new The residual matrix of sweat volume after removing the influence of the current components; (5315) Repeat steps (5311)-(5314) until p principal components are extracted; (532) Calculate the regression coefficient β, and connect the sweat volume characteristic X with the sweat volume Y, as shown in the following expression: β=W(P T W) -1 c The PLSR model at this time is: Where ∈ represents the error term, and thus the sweat volume characteristic X and the predicted sweat volume value are established. The mapping relationship between them.

7. The sweat volume monitoring system based on RFID passive multi-frequency sensing according to claim 6, characterized in that, Step (6) specifically includes: (61) Regarding the test data X text Standardization processing was performed to obtain standardized sweat volume characteristic test data X. test,std The expression is as follows: (62) Using the regression coefficient β for prediction, the prediction results are obtained. The expression is as follows: (63) The prediction results are denormalized to obtain the final predicted value of sweat volume. The expression is as follows:

8. A method for monitoring sweat volume based on RFID passive multi-frequency sensing, based on the system described in any one of claims 1-7, characterized in that, The steps are as follows: 1) Attach the passive sweat volume sensor to the target location on the human body; 2) The sweat volume sensing all-in-one machine continuously transmits multi-frequency wireless radio frequency signals and collects the wireless radio frequency signals backscattered by the passive sweat volume sensor in real time, and sends them to the data processing platform. 3) The data processing platform calculates the phase information of the multi-frequency signals from the above-mentioned radio frequency signals; 4) Eliminate the influence of human movement on the perception results by using dual RFID tags and multi-frequency signal phase difference method to obtain sweat volume characteristics; 5) Calculate all sweat volume characteristics within the measurement range and train the sweat volume regressor; 6) Input the sweat volume characteristic test data into the sweat volume regressor for prediction to obtain the sweat volume.

Images