A wearable urinary incontinence monitoring system based on multi-modal sensor fusion

Abstract

The application discloses a wearable urinary incontinence monitoring system based on multi-modal sensor fusion and relates to the technical field of urinary incontinence monitoring, and solves the problems that the existing urinary incontinence monitoring cannot be carried out in daily life and has poor monitoring precision, and the scheme points comprise a sensing diaper, a signal acquisition front end, a user mobile terminal and a cloud management platform; the sensing diaper is used for collecting user urine leakage signals through a sensor; the signal acquisition front end is used for collecting user activity signals through the sensor, pre-processing the user urine leakage signals and the user activity signals to obtain user data and forwarding; the user mobile terminal is used for carrying out urine leakage frequency monitoring and user behavior activity monitoring through the user data to obtain a monitoring report, and uploading the monitoring report and the user data to the cloud management platform; the cloud management platform is used for establishing an electronic case through the monitoring report and the user data, and delivering analysis results to the user mobile terminal; and the monitoring precision can be improved through temperature data and pH data.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of urinary incontinence monitoring, more particularly, it relates to a wearable urinary incontinence monitoring system based on multi-modal sensor fusion. BACKGROUND

[0002] Female stress urinary incontinence (SUI) is a high-incidence female pelvic floor disease and the most common type of urinary incontinence. SUI is defined as "a clinical symptom of involuntary urine loss during coughing, sneezing, jumping and physical activity." The detection of the degree of SUI is divided into subjective evaluation and objective examination. The subjective evaluation has strong subjectivity and cannot accurately reflect the urine volume and urine frequency; the objective examination can accurately reflect the urine volume, but the process is complicated. Both methods need to be carried out in a hospital environment, the detection time is limited, and the user's urinary incontinence in daily life cannot be monitored, which is not convenient for doctors to timely grasp the change of the user's urinary incontinence degree.

[0003] At present, there are intelligent urine sheets on the market that can be used in daily life, but they mainly remind of urine leakage or urinary incontinence, cannot realize online or offline urine leakage frequency and urine leakage volume monitoring, and cannot monitor and analyze the influence of user behavior activities (such as coughing, sneezing, jumping and physical labor activities that change abdominal pressure) on urinary incontinence; and for urinary incontinence monitoring itself, it is also easily affected by the reverse seepage of urine in the urine sheet core layer and the user's urethral secretions, so the monitoring accuracy is not high.

[0004] Therefore, the present application provides a wearable urinary incontinence monitoring system based on multi-modal sensor fusion to solve the above problems. SUMMARY

[0005] The purpose of the present application is to provide a wearable urinary incontinence monitoring system based on multi-modal sensor fusion to solve the problems of existing urinary incontinence monitoring that cannot be carried out in daily life and poor monitoring accuracy.

[0006] The above technical purpose of the application is realized by the following technical scheme: comprising: a sensing diaper, a signal acquisition front end, a user mobile terminal and a cloud management platform; the sensing diaper is used to collect user urine leakage signals through sensors, including: humidity signals, temperature signals and pH signals; the signal acquisition front end is used to collect user activity signals through sensors, including: pressure signals and inertia signals, and receives user urine leakage signals of the sensing diaper, pre-processes the user urine leakage signals and the user activity signals to obtain user data, and forwards the user data to the user mobile terminal; the user mobile terminal is used to monitor urine leakage frequency and emergency pressure through user data to obtain a monitoring report, and uploads the monitoring report and the user data to the cloud management platform; the cloud management platform is used to establish an electronic case through the monitoring report and the user data, and feeds back analysis results to doctors, and feeds back data to the user mobile terminal according to needs; wherein the monitoring report includes: the number of urinary incontinence and user behavior activities, the number of urinary incontinence is obtained based on humidity data, and is corrected based on temperature data and pH data; the user behavior activities are obtained based on pressure data and inertia data.

[0007] By the above technical scheme, physiological signals and activity signals of users in daily life are collected through the sensing diaper, the sensor signals are preliminarily processed through the signal acquisition front end to obtain user data and are forwarded, the user data are analyzed through the user mobile terminal to realize real-time monitoring and reporting of urine leakage frequency and user behavior activities, and the cloud management platform is used to track changes of user conditions. By monitoring urinary incontinence of users in daily life through the sensing diaper, the influence of reverse water seepage of the diaper and urethral secretions on the monitoring result can be effectively avoided through cooperation of humidity data, temperature data and pH data, and the monitoring accuracy is improved.

[0008] In a possible implementation, the user mobile terminal comprises a urine incontinence frequency analysis module, which comprises: a humidity data acquisition module, configured to construct humidity change sequence data of humidity change over time based on humidity data; a humidity data processing module, configured to perform low-pass filtering and standardization processing on the humidity change sequence data to obtain standard humidity change data; an extreme point query module, configured to calculate a first derivative of the standard humidity change data, and query zero-crossing points from the first derivative, the standard humidity change data corresponding to the zero-crossing points being extreme point data; an extreme point screening module, configured to set a humidity change threshold, and screen data points exceeding the set humidity change threshold from the extreme point data corresponding to all the zero-crossing points as candidate wave peak points; a urine incontinence frequency analysis module, configured to set a time window at each candidate wave peak point, check whether the humidity change data in the time window meets a urine leakage feature, and record a number of candidate wave peak points meeting the urine leakage feature as a urine incontinence frequency; a urine incontinence frequency correction module, configured to correct the urine incontinence frequency once by using humidity data, correct the urine incontinence frequency twice by using temperature data, and obtain a urine incontinence frequency excluding diaper reverse water seepage interference, and correct the urine incontinence frequency thrice by using pH data, and obtain a urine incontinence frequency excluding urethral secretion interference.

[0009] In a possible implementation, the urine incontinence frequency correction module comprises a first correction module, configured to perform the following steps: obtaining a humidity change amplitude and a humidity peak value recovery time of normal urine leakage, obtaining a humidity change amplitude and a humidity peak value recovery time of each urine incontinence time, and regarding urine incontinence frequencies with a humidity change amplitude lower than the humidity change amplitude of normal urine leakage and a humidity peak value recovery time longer than the humidity peak value recovery time of normal urine leakage as error frequencies caused by reverse osmosis.

[0010] In a possible implementation, the urine incontinence frequency correction module comprises a second correction module, configured to perform the following steps: constructing temperature waveform data of temperature change over time based on temperature data; extracting a temperature of each urine incontinence time from the temperature waveform data, regarding urine incontinence frequencies with a temperature lower than a normal urine temperature as error frequencies caused by reverse osmosis, and excluding the error frequencies to obtain a urine incontinence frequency excluding diaper reverse water seepage interference.

[0011] In a possible implementation, the number of incontinence correction module includes a three-step correction module, configured to perform the following steps: based on the pH data, constructing pH change data of the pH value over time; performing Kalman filtering on the pH change data to obtain standard pH data; extracting a pH feature value of each incontinence time from the standard pH data, the pH feature value being one of an average value, a standard deviation, a maximum value, and a minimum value; setting a urine pH threshold and a secretion pH threshold, and regarding an incontinence number of the pH feature value of the incontinence time being located outside the urine pH threshold interval and inside the secretion pH threshold interval as an error number caused by urethral secretion, and excluding the error number to obtain an incontinence number excluding the interference of the urethral secretion.

[0012] In a possible implementation, the extreme point screening module is further configured to: set the humidity change threshold to decrease with an increase in the use time of the diaper.

[0013] In a possible implementation, the humidity change threshold calculation formula is:

[0014]

[0015] wherein θ(t) is the humidity change threshold at the current time t, θ0 is an initial humidity change threshold, λ is a decay rate, and t0 is a time at which the humidity sensor starts to work.

[0016] In a possible implementation, the user mobile terminal includes a user behavior activity analysis module, configured to perform the following steps: based on the pressure data and the inertial data, constructing pressure sequence data and inertial force sequence data of the pressure and the inertial force over time; performing time synchronization processing and low-pass filtering processing on the pressure sequence data and the inertial force sequence data; extracting feature values of the pressure sequence data, including a peak value, an average value, and a standard deviation, and extracting feature values of the inertial force sequence data, including an amplitude value, a direction change, and a frequency, and splicing the feature values of the pressure sequence data and the feature values of the inertial force sequence data to form a multi-dimensional feature vector; and sending the multi-dimensional feature vector to a trained motion recognition model to obtain a user behavior activity type.

[0017] In a possible implementation, the user mobile terminal includes an association model analysis module, configured to perform the following steps: when detecting an incontinence, detecting whether a user behavior activity exists within T1 seconds before the incontinence, and when the user behavior activity exists, recording an incontinence time and a user behavior activity time; when detecting a user behavior activity, detecting whether an incontinence exists within T2 seconds after the user behavior activity, and when the incontinence exists, recording the incontinence time and the user behavior activity time; and based on the incontinence time and the user behavior activity time, establishing a time association model of the incontinence and the user behavior activity.

[0018] In a possible implementation, the user mobile terminal comprises a urine leakage amount analysis module, configured to perform the following steps: when the sensing diaper adopts a capacitive humidity sensor, according to experimental analysis of the capacitance change corresponding to different urine leakage amounts, the urine leakage amount is obtained according to the capacitance change before and after each urinary incontinence; when the sensing diaper adopts a resistive humidity sensor, the urine leakage amount is obtained according to the area enclosed by the resistance change curve before and after each urinary incontinence.

[0019] Compared with the prior art, the application has the following beneficial effects: first, a multi-modal sensor fusion urinary incontinence monitoring system is provided, which can accurately detect the urine leakage frequency and user behavior activities (such as coughing, laughing, jumping, etc.) in the user's daily life, and realize mobile terminal monitoring and cloud remote analysis; second, the urine leakage frequency and user behavior activities are obtained, which can analyze the correlation between the two and provide a basis for clinical data analysis; third, based on the urinary incontinence frequency obtained based on humidity data, the urinary incontinence frequency is corrected by temperature data and pH data, which can effectively avoid the influence of diaper reverse water penetration and urethral secretions on the monitoring results. BRIEF DESCRIPTION OF DRAWINGS

[0020] The accompanying drawings, which are included to provide a further understanding of the embodiments of the application and constitute a part of this application, illustrate embodiments of the application and are used to explain the embodiments of the application. In the drawings:

[0021] Figure 1 FIG. 1 is a structural schematic diagram of a wearable urinary incontinence monitoring system based on multi-modal sensor fusion;

[0022] Figure 2 FIG. 5 is a layout schematic diagram of a humidity sensor, a temperature sensor, and a pH sensor;

[0023] Figure 3 FIG. 6 is a layout schematic diagram of a pressure sensor and an inertial sensor. DETAILED DESCRIPTION

[0024] In order to make the purpose, technical scheme and advantages of the application more clear, the application is further described in detail below in combination with embodiments and drawings. The illustrative embodiments of the application and the description thereof are used to explain the application, and do not limit the application.

[0025] Please refer to Figure 1 as shown, Figure 1A structural schematic diagram of a wearable urinary incontinence monitoring system based on multi-modal sensor fusion. It includes a sensing diaper, a signal acquisition front end, a user mobile terminal and a cloud management platform; the sensing diaper is used to collect user urine leakage signals through sensors, including humidity signals, temperature signals and pH signals; the signal acquisition front end is used to collect user activity signals through sensors, including pressure signals and inertial signals, and receives user urine leakage signals from the sensing diaper, pre-processes the user urine leakage signals and the user activity signals to obtain user data, and forwards the user data to the user mobile terminal; the cloud management platform is used to establish an electronic case through the monitoring report and the user data, and to issue analysis results to doctors, and to feed back data to the user mobile terminal as needed; wherein the monitoring report includes the number of urinary incontinence and user behavior activities, the number of urinary incontinence is obtained based on humidity data and corrected based on temperature data and pH data; the user behavior activities are obtained based on pressure data and inertial data.

[0026] Specifically, the wearable urinary incontinence monitoring system is composed of a sensing diaper, a signal acquisition front end, a user mobile terminal, and a cloud management platform. The functions and principles of each module are as follows: the sensing diaper is used to collect user urine leakage signals, and is embedded with multi-modal sensors, including a humidity sensor, a temperature sensor, and a pH sensor. The humidity sensor can monitor the humidity changes caused by urine dripping and reflect the number of urinary incontinence; the temperature sensor can monitor the temperature of the urine and reflect the generation time of the urine; and the pH sensor can perform in-depth analysis on the pH value of the incontinence body fluid. The signal acquisition front end can be used to collect user activity signals, and is provided with a pressure sensor and an inertial sensor. The pressure sensor can monitor the user's abdominal force, and the inertial sensor can monitor the user's acceleration, together reflecting the user's behavior activity and the intensity of the user's behavior activity. The signal acquisition front end can also be used to preprocess the user urine leakage signals and the user activity signals to obtain user data. The signal acquisition front end is integrated with a low-power processor and connected with the sensing diaper to preprocess the user urine leakage signals and the user activity signals and obtain user data for subsequent processing. The signal acquisition front end is also provided with a Bluetooth or WiFi chip to forward the user data. The user mobile terminal can be an APP installed on the user's mobile phone and connected with the signal acquisition front end through wireless transmission to receive and store the user data. The user mobile terminal is internally integrated with a local algorithm to realize real-time monitoring of the urine leakage frequency and the user behavior activity through analysis of the user data. In combination with an expert knowledge base, the user mobile terminal can provide a user-readable monitoring report. The user data and the preliminary analysis monitoring report can be uploaded to a cloud server through WiFi or mobile data for further viewing and analysis by a doctor. The cloud management platform can be used by the doctor to view the user data and the monitoring report. The doctor can intuitively understand the user's condition changes through a data visualization tool provided by the platform, and can also analyze the condition based on the user data and the monitoring report. The doctor can decide whether to provide the analysis result to the APP of the user's mobile phone to realize personalized management. In order to better track the user's condition changes, the doctor can also establish a complete patient case system on the cloud management platform based on the above data, establish an independent case file for each user, and update the user's condition changes and the doctor's suggestions and other information in real time.

[0027] The improvements of the present scheme are as follows: first, a multi-modal sensor fusion urinary incontinence monitoring system is provided, which can accurately detect the urine leakage frequency and the user behavior activity (such as coughing, laughing, jumping, etc.) in the user's daily life, and realize mobile terminal monitoring and cloud remote analysis; second, the user urine leakage frequency and the user behavior activity are obtained, and the correlation between the two can be analyzed to provide a basis for clinical data analysis; and third, based on the number of urinary incontinence obtained based on the humidity data, the number of urinary incontinence is corrected through temperature data and pH data, which can effectively avoid the influence of diaper reverse water seepage and urethral secretions on the monitoring result.

[0028] Regarding the sensing diaper and the signal acquisition front end: the sensing diaper can be arranged in a disposable sheet structure in the crotch of the underpants, and the signal acquisition front end can be arranged in a reusable box structure in the abdomen of the underpants, and the sensing diaper and the signal acquisition front end are connected through a connector or a wire. The sensing diaper is a multi-layer structure, at least including: a surface layer, a sensing layer and a flow guiding and absorbing layer, the surface layer is in contact with the human body, the sensing layer is arranged with a multi-modal sensor, and the flow guiding and absorbing layer guides and absorbs urine. The sensing layer can be a single-layer or multi-layer structure, and the sensing layer is arranged with a multi-modal sensor; the multi-modal sensor can be made of flexible conductive yarn, pH sensing yarn (or micro pH sensing chip), temperature sensing yarn (or patch particle), etc., and can be set in a surface electrode + core electrode manner, the electrodes therebetween can be arranged in an array manner, and can be fixed by using adhesives, printing, embroidery or hot pressing, etc., and can use capacitive electrodes, resistive electrodes, battery electrodes, biochemical electrodes, temperature sensors, etc. It should be noted that the shape of the sensing diaper is adapted to the crotch of the underpants, and is fixed to the side of the crotch of the underpants close to the human body for monitoring the leakage condition; wherein the sensing layer is provided with humidity sensors, temperature sensors and pH sensors, which are densely arranged in an array, please refer to Figure 2 The signal acquisition front end is a box body, which is fixed to the side of the abdomen of the underpants away from the human body for monitoring the pressure and inertial force caused by behavior; wherein the pressure sensor is fixed to the outside surface of the bottom of the box body, i.e. the side contacting the abdomen, and the inertial sensor is fixed to the inside of the box body, please refer to Figure 3 .

[0029] The user mobile terminal is one of the focuses of the present application. In one possible implementation, the user mobile terminal includes a urinary incontinence frequency analysis module, which includes: a humidity data acquisition module for constructing humidity change sequence data based on humidity data, the humidity change sequence data being humidity change over time; a humidity data processing module for low-pass filtering and standardizing the humidity change sequence data to obtain standard humidity change data; an extreme point query module for calculating the first derivative of the standard humidity change data, querying zero-crossing points from the first derivative, and the standard humidity change data corresponding to the zero-crossing points being extreme point data; an extreme point screening module for setting a humidity change threshold and screening data points exceeding the set humidity change threshold from all extreme point data corresponding to the zero-crossing points as candidate wave peak points; a urinary incontinence frequency analysis module for setting a time window at each candidate wave peak point and checking whether the humidity change data in the time window meets the leakage characteristics, and recording the number of candidate wave peak points meeting the leakage characteristics as the urinary incontinence frequency; a urinary incontinence frequency correction module for correcting the urinary incontinence frequency once by humidity data, correcting the urinary incontinence frequency twice by temperature data to obtain the urinary incontinence frequency excluding the interference of reverse water seepage of the diaper, and correcting the urinary incontinence frequency three times by pH data to obtain the urinary incontinence frequency excluding the interference of urethral secretions.

[0030] Specifically, the number of urinary incontinence can be obtained by peak detection, by monitoring the peaks of humidity sensor data to identify the occurrence of urine leakage events. The specific algorithm process is as follows:

[0031] 1. Data collection. Collect real-time humidity signals through humidity sensors, process humidity signals into humidity data that can be received by user mobile terminals through signal collection front ends, and form humidity change sequence data based on humidity data.

[0032] 2. Data preprocessing. Apply a low-pass filter to the humidity change sequence data for denoising, and then standardize to ensure data consistency, obtaining standard humidity change data.

[0033] 3. Feature extraction. Calculate the first derivative of the standard humidity change data to identify the rate of change of the data, and find the zero-crossing points in the first derivative signal. These points correspond to the extreme points (peaks and valleys) of the original humidity data.

[0034] 4. Peak identification. Set a humidity change threshold, only when the humidity data changes beyond this threshold, it is considered that a urine leakage event may have occurred, so filter out the points where the humidity change exceeds the set humidity change threshold from all zero-crossing points as candidate peak points. The calculation formula is:

[0035]

[0036] Where w is the size of the sliding window, and threshold is the set threshold.

[0037] 5. Urine leakage event confirmation. Set a time window around each candidate peak point, and check whether the humidity change in the time window meets the characteristics of urine leakage (such as rapid rise, peak reached in a short time). The formula is:

[0038] E={t i ∈P|t i -t i-1 >m interval}(2-2)

[0039] Where m interval is the minimum time interval between peaks.

[0040] 6. Urine leakage event correction. Further verification is made in combination with other parameters (such as humidity recovery time, temperature data, pH data, etc.) to ensure accurate identification of urine leakage events.

[0041] It should be noted that this method can quickly and accurately capture the occurrence of urine leakage events, and has higher real-time performance, sensitivity and portability compared to offline hospital urine pad tests.

[0042] Regarding peak recognition, in one possible implementation, dynamic threshold monitoring can be employed. The extreme point screening module is further configured to set the humidity change threshold to decrease with the increase of the usage time of the diaper to compensate for the decrease of the sensitivity of the humidity sensor with the increase of the usage time. The humidity change threshold is calculated as follows:

[0043]

[0044] wherein θ(t) is the humidity change threshold at current time t, θ0 is the initial humidity change threshold, λ is the decay rate, and t0 is the time when the humidity sensor starts to work.

[0045] It should be noted that the conventional threshold detection algorithm employs a fixed threshold, which cannot dynamically change with the actual situation during use, resulting in a gradual increase in error over a long period of use.

[0046] Regarding the correction of urine leakage events, in one possible implementation, the urine incontinence frequency correction module includes a primary correction module configured to perform the following steps: obtaining the humidity change amplitude and humidity peak recovery time of normal urine leakage, obtaining the humidity change amplitude and humidity peak recovery time of each urine incontinence time, and taking the urine incontinence times with a humidity change amplitude lower than the normal urine leakage humidity change amplitude and a humidity peak recovery time longer than the normal urine leakage humidity peak recovery time as the error times caused by reverse osmosis.

[0047] Specifically, based on the humidity change amplitude and humidity peak recovery time of normal urine leakage collected, the urine leakage false detection caused by reverse osmosis is preliminarily excluded. The reason is that the humidity change of normal urine leakage is more rapid and the humidity peak recovery is faster than that of reverse osmosis. The slope of the humidity waveform and the recovery time after reaching the peak value can be analyzed by threshold judgment and machine learning method to distinguish normal urine leakage from reverse osmosis.

[0048] In one possible implementation, the urine incontinence frequency correction module includes a secondary correction module configured to perform the following steps: based on the temperature data, constructing temperature waveform data of temperature change over time; extracting the temperature at each urine incontinence time from the temperature waveform data; taking the urine incontinence times with a temperature lower than the normal urine temperature as the error times caused by reverse osmosis; and excluding the error times to obtain the urine incontinence times excluding the interference of reverse osmosis of the diaper.

[0049] Specifically, by the normal leakage urine temperature and the reverse osmosis urine temperature, the leakage urine misjudgment caused by reverse osmosis is preliminarily excluded. The reason is that the temperature of urine just after dropping is higher than the temperature of urine after reverse osmosis. Therefore, the leakage urine collection times lower than the normal urine temperature can be defined as the reverse osmosis times. The temperature sensor is used to collect data in real time, form a waveform of temperature change, and distinguish the leakage urine and the reverse osmosis through threshold judgment and machine learning according to the temperature change rate.

[0050] In a possible implementation, the urine incontinence times correction module includes a three-time correction module, configured to perform the following steps: based on the pH data, constructing pH change data of the pH value over time; performing Kalman filtering on the pH change data to obtain standard pH data; extracting a pH feature value of each urine incontinence time from the standard pH data, the pH feature value being one of an average value, a standard deviation, a maximum value, and a minimum value; setting a urine pH threshold and a secretion pH threshold, and regarding the urine incontinence times, whose pH feature value is located outside the urine pH threshold interval and located within the secretion pH threshold interval, as error times caused by urethral secretions, and excluding the error times to obtain the urine incontinence times excluding the interference of the urethral secretions.

[0051] Specifically, the pH sensor integrated in the sensing diaper is used to monitor the pH values of urine and secretions in real time. Unlike the humidity sensor which can only monitor the presence of liquid, the pH sensor can distinguish between urine and other body fluid secretions, such as sweat or vaginal secretions. By using the different characteristics of pH values, urine incontinence events and other secretion events can be distinguished. The specific algorithm process is as follows: 1. Data collection. The real-time pH signal is collected by the pH sensor, and the pH signal is processed into pH data that can be received by the user mobile terminal through the signal collection front end, and the pH change data is formed based on the pH data. 2. Data preprocessing. The Kalman filtering method is used to remove noise and calibrate the pH sensor to ensure data accuracy and obtain standard pH data. 3. Feature extraction. The features of the pH data in the urine incontinence time interval are extracted, such as the average value, the standard deviation, the maximum value, and the minimum value. 4. Event identification. The urine pH threshold and the secretion pH threshold are set, and the urine incontinence events and other secretion events are identified by comparing the real-time pH value with the threshold.

[0052] In addition, the frequency of urinary incontinence can also be monitored in combination with image analysis. Signal image analysis technology is used to distinguish between normal physiological changes and urinary leakage events. The specific algorithm process is: 1. Data acquisition, the same as above. 2. Data preprocessing, the same as above. 3. Signal conversion to image. Convert time series data into image representation (such as plotting the signal amplitude into a two-dimensional image). 4. Feature extraction. Extract features from the image, such as edges, textures, shapes, etc. 5. Pattern recognition. Use machine learning technology to classify the extracted features to distinguish between normal physiological changes and urinary leakage events. Train classification models (such as random forests) and make predictions for new data. 6. Urinary leakage event detection. Identify urinary leakage events based on the classification results and record relevant information. It should be noted that traditional methods cannot effectively distinguish between normal physiological changes and urinary leakage events, while signal image analysis can improve the accuracy of detection through image pattern recognition technology.

[0053] The frequency of urinary incontinence can also be monitored by combining it with machine learning algorithms. The specific algorithm process is as follows: 1. Data acquisition, as described above. 2. Data preprocessing, as described above. 3. Feature extraction. Extract time domain features, such as mean, standard deviation, maximum, minimum, etc.; extract frequency domain features, such as spectral features after Fourier transform; extract time series features, such as autocorrelation coefficient, sliding window statistical features, etc. 4. Data labeling. Manual labeling is used to manually identify urine leakage events and label the data set. Automatic labeling can also be used, combining traditional algorithms (such as peak detection) for preliminary labeling, which is then manually reviewed. 5. Model selection and training. Select an appropriate machine learning model based on the data characteristics, such as random forest, support vector machine (SVM), long short-term memory network (LSTM), etc., train the model using the labeled data set, and adjust the model parameters to improve accuracy and generalization ability. 6. Model validation and evaluation. Evaluate model performance through cross-validation to avoid overfitting, and use indicators such as accuracy, recall rate, and F1 score to evaluate model effectiveness. It should be noted that compared with traditional threshold judgment methods, machine learning algorithms can improve the recognition accuracy and adaptability of complex urine leakage patterns through large amounts of data training.

[0054] In one possible implementation, the user mobile terminal includes a user behavior activity analysis module, which is used to perform the following steps: based on pressure data and inertia data, construct pressure sequence data and inertia force sequence data of pressure and inertia force changing over time; perform time synchronization processing and low-pass filtering on the pressure sequence data and inertia force sequence data; extract the characteristic values ​​of the pressure sequence data, including: peak value, average value, and standard deviation, and extract the characteristic values ​​of the inertia force sequence data, including: amplitude, direction change, and frequency; splice the characteristic values ​​of the pressure sequence data and the characteristic values ​​of the inertia force sequence data to form a multidimensional feature vector; and send the multidimensional feature vector into a trained action recognition model to obtain the user behavior activity type.

[0055] Specifically, the user's stress data and inertial data are monitored through the pressure sensor and the IMU accelerometer (inertial sensor) to distinguish between coughing, laughing, jumping, and other behaviors. Traditional motion monitoring usually only uses the IMU accelerometer or the pressure sensor, while the present scheme innovatively proposes to integrate the two, which can more comprehensively and accurately capture the user's motion situation. The specific algorithm process is as follows: 1. Data collection. Collect the pressure data and inertial data collected by the pressure sensor and the IMU accelerometer in real time, and convert the pressure data and the inertial data into pressure sequence data and inertial force sequence data that change over time. 2. Data preprocessing. Synchronize the pressure sequence data and the inertial force sequence data to ensure that their time axes are aligned, and use a low-pass filter to remove noise. 3. Feature extraction. Extract the peak value, average value, and standard deviation of the pressure sequence data, and extract the amplitude, direction change, and frequency of the inertial force sequence data, and combine the pressure and inertial force features to form a multi-dimensional feature vector. 4. Pattern recognition. For example, threshold method: set an empirical threshold to quickly identify obvious motion events; for example, machine learning model: train a machine learning model (such as random forest, SVM, neural network) to recognize complex motion patterns, use a labeled data set for training, the model input is the joint feature vector, and the output is the motion type (such as coughing, laughing, jumping, etc.).

[0056] In addition, machine learning algorithms can also be combined: through training data, accurately identify different degrees of user behavior activities, and judge the possibility and severity of urine leakage triggering. Innovatively, the user behavior activities are classified into grades, corresponding to the urine leakage situation, and the user behavior activities are associated with the urine leakage degree. Machine learning algorithms can more finely classify and judge the severity of different user behavior activities. The specific algorithm process is similar to the machine learning algorithm in the urine incontinence frequency monitoring.

[0057] In one possible implementation, the user mobile terminal includes an association model analysis module for performing the following steps: when urine incontinence is detected, it is detected whether there is user behavior activity within T1 seconds before the urine incontinence occurs, and when there is user behavior activity, the urine incontinence time and the user behavior activity time are recorded; when user behavior activity is detected, it is detected whether there is urine incontinence within T2 seconds after the user behavior activity occurs, and when there is urine incontinence, the urine incontinence time and the user behavior activity time are recorded; based on the urine incontinence time and the user behavior activity time, an association model of urine incontinence and user behavior activity in time is established.

[0058] Specifically, the incontinence and user behavior activities are determined simultaneously: when the urine leakage event occurs, the user behavior activities (such as coughing, laughing, jumping, etc.) in the previous two seconds are traced back, and whether the user behavior activities cause the urine leakage event can be analyzed through the user behavior activities. The determination of the user behavior activities causing the urine leakage: when the user behavior activities (such as coughing, laughing, jumping, etc.) are detected, it is continuously monitored and recorded whether the urine leakage event occurs within two seconds after the user behavior activities occur. In this way, the time correlation between the user behavior activities and the urine leakage event can be established, so as to evaluate the influence of the actions on the urine leakage.

[0059] In a possible implementation, the user mobile terminal includes a urine leakage amount analysis module for performing the following steps: when the sensing urine sheet adopts a capacitive humidity sensor, the amount of urine leakage is obtained according to the change amount of the capacitance corresponding to different urine leakage amounts according to the experimental analysis, and the amount of urine leakage is obtained according to the change amount of the capacitance before and after each incontinence; when the sensing urine sheet adopts a resistance humidity sensor, the amount of urine leakage is obtained according to the area enclosed by the resistance change curve before and after each incontinence.

[0060] Specifically, 1) when the sensing urine sheet adopts a capacitive humidity sensor, different electrode data of different precision levels corresponding to different capacitive materials are selected by changing the electrode spacing and electrode size of the capacitive sensor, and a corresponding model of the amount of urine leakage and the change of the capacitance is established according to the experiment. The amount of urine leakage is analyzed according to the change amount of the capacitance after different urine leakages, the urine leakage frequency is analyzed according to the change value of the capacitance each time. 2) When the sensing urine sheet adopts a resistance humidity sensor, the amount of urine is measured according to the area enclosed by the resistance change curve before and after the incontinence, and the amount of urine leakage is determined. The amount of urine leakage is mainly determined according to the resistance change waveform, the sampling amplitude and the sampling points and time required for each sampling amplitude change are determined according to the wave crest and trough of the waveform, and finally the amount of urine leakage corresponding to different urine amounts is analyzed according to the sampling amplitude, sampling points and time under different urine amounts. The urine leakage frequency is determined according to the change of the sampling amplitude each time, and the accuracy of the urine leakage frequency and urine amount monitoring is greatly improved through the analysis of the collected data.

[0061] The system aims to realize the precise monitoring of the user's daily urinary incontinence frequency and the user's behavior activities before and after incontinence by combining sensing diapers, signal acquisition front end, user mobile terminal and cloud management platform. Among them, the measurement of urinary incontinence frequency is based on high-sensitivity humidity data, and the accuracy can reach 90%; the detection of urine leakage is through the construction of a high-precision mathematical model, and the average absolute error is not higher than 5mL. The monitoring of user behavior activities covers actions such as standing up, sitting down, walking, running and falling, especially the fall detection needs to distinguish the front, back, left and right four directions, and the accuracy can reach more than 95%. The system uses wireless Bluetooth transmission and is equipped with local storage function to ensure data integrity and accuracy, and the data is highly reliable and portable. In addition, the front-end module design focuses on wearing comfort, small size and light weight, improving the user experience. Low-power design ensures continuous working time of more than 16 hours, except for night rest, to meet the long-term monitoring needs.

[0062] The above specific embodiments further illustrate the purpose, technical solutions and beneficial effects of the present application. It should be understood that the above description is only a specific embodiment of the present application and is not intended to limit the protection scope of the present application. Any modification, equivalent replacement, improvement, etc. within the spirit and principles of the present application shall be included in the protection scope of the present application.

Claims

1. A wearable urinary incontinence monitoring system based on multi-modal sensor fusion, characterized in that, The application relates to a urine sensing system, comprising a sensing diaper, a signal acquisition front end, a user mobile terminal and a cloud management platform. The sensing diaper is used for collecting user urine leakage signals through sensors, and the sensing diaper comprises humidity signals, temperature signals and pH signals. The signal acquisition front end is used for collecting user activity signals through sensors, and the signal acquisition front end comprises pressure signals and inertia signals, and receives the user urine leakage signals of the sensing diaper, pre-processes the user urine leakage signals and the user activity signals to obtain user data, and forwards the user data to the user mobile terminal. The user mobile terminal is used for monitoring urine leakage frequency and user behavior activity through the user data to obtain a monitoring report, and uploads the monitoring report and the user data to the cloud management platform. The cloud management platform is used for establishing an electronic case through the monitoring report and the user data, and feeds back analysis results to doctors and the user mobile terminal as required. The monitoring report comprises urine incontinence times and user behavior activity, the urine incontinence times are obtained based on humidity data and are corrected based on temperature data and pH data, and the user behavior activity is obtained based on pressure data and inertia data. The user mobile terminal comprises a correlation model analysis module, which is used for performing the following steps: When urine incontinence is detected, whether user behavior activity exists within T1 seconds before urine incontinence is detected is detected, when the user behavior activity exists, urine incontinence time and user behavior activity time are recorded; When user behavior activity is detected, whether urine incontinence exists within T2 seconds after the user behavior activity is detected is detected, when the urine incontinence exists, urine incontinence time and user behavior activity time are recorded; Based on the urine incontinence time and the user behavior activity time, a correlation model of urine incontinence and user behavior activity in time is established. The user mobile terminal comprises a urine incontinence times analysis module, and the urine incontinence times analysis module comprises:

2. The wearable urinary incontinence monitoring system based on multi-modal sensor fusion as claimed in claim 1, wherein, A humidity data acquisition module is used for constructing humidity change sequence data of humidity change with time based on humidity data; A humidity data processing module is used for performing low-pass filtering and standardization processing on the humidity change sequence data to obtain standard humidity change data; An extreme point query module is used for calculating a first derivative of the standard humidity change data, and querying zero-crossing points from the first derivative, and the standard humidity change data corresponding to the zero-crossing points is extreme point data; An extreme point screening module is used for setting a humidity change threshold, and screening data points exceeding the set humidity change threshold from the extreme point data corresponding to all zero-crossing points as candidate wave peak points; A urine incontinence times analysis module is used for setting a time window at each candidate wave peak point, checking whether humidity change data in the time window meets urine leakage characteristics, and recording the number of candidate wave peak points meeting the urine leakage characteristics as urine incontinence times; A urine incontinence times correction module is used for correcting the urine incontinence times once through humidity data, correcting the urine incontinence times twice through temperature data to obtain urine incontinence times excluding reverse water seepage interference of the sensing diaper, and correcting the urine incontinence times three times through pH data to obtain urine incontinence times excluding urethral secretion interference. ​ 3. The wearable urinary incontinence monitoring system based on multi-modal sensor fusion as claimed in claim 2, wherein, The urine loss frequency correction module comprises a first correction module, configured to perform the following steps: Obtain the humidity change amplitude and humidity peak value recovery time of normal urine leakage, obtain the humidity change amplitude and humidity peak value recovery time of each urine loss time, and regard the urine loss times with a humidity change amplitude lower than the humidity change amplitude of normal urine leakage and a humidity peak value recovery time longer than the humidity peak value recovery time of normal urine leakage as error times caused by reverse osmosis.

4. The wearable urinary incontinence monitoring system based on multi-modal sensor fusion as claimed in claim 2, wherein, The urine loss frequency correction module comprises a second correction module, configured to perform the following steps: Based on the temperature data, construct temperature waveform data of temperature change over time; Extract the temperature of each urine loss time from the temperature waveform data, and regard the urine loss times with a temperature lower than the normal urine temperature as error times caused by reverse osmosis, to obtain urine loss times excluding the interference of reverse osmosis of diaper water.

5. The wearable urinary incontinence monitoring system based on multi-modal sensor fusion as claimed in claim 2, wherein, The urine loss frequency correction module comprises a third correction module, configured to perform the following steps: Based on the pH data, construct pH change data of pH value change over time; Perform Kalman filtering processing on the pH change data to obtain standard pH data; Extract the pH characteristic value of each urine loss time from the standard pH data, the pH characteristic value being one of mean value, standard deviation, maximum value and minimum value; Set urine pH threshold value and secretion pH threshold value, regard the urine loss times with a pH characteristic value located outside the urine pH threshold value interval and located within the secretion pH threshold value interval as error times caused by urethral secretion, and exclude the error times to obtain urine loss times excluding the interference of urethral secretion.

6. The wearable urinary incontinence monitoring system based on multi-modal sensor fusion as claimed in claim 2, wherein, The extreme point screening module is further configured to set the humidity change threshold value to decrease with the increase of diaper use time.

7. The wearable urinary incontinence monitoring system based on multi-modal sensor fusion as claimed in claim 6, wherein, The humidity change threshold value calculation formula is: wherein is a humidity change threshold for the current time t is a humidity change threshold for the current time is an initial humidity change threshold, is a decay rate, is a time at which the humidity sensor starts to work.

8. The wearable urinary incontinence monitoring system based on multi-modal sensor fusion as claimed in claim 1, wherein, The user mobile terminal comprises a user behavior activity analysis module, configured to perform the following steps: Based on the pressure data and inertial data, construct pressure sequence data and inertial force sequence data of pressure and inertial force change over time; Perform time synchronization processing and low-pass filtering processing on the pressure sequence data and inertial force sequence data; Extract the characteristic values of the pressure sequence data, including peak value, mean value and standard deviation, and extract the characteristic values of the inertial force sequence data, including amplitude, direction change and frequency, and splice the characteristic values of the pressure sequence data and the characteristic values of the inertial force sequence data to form a multi-dimensional feature vector; Send the multi-dimensional feature vector to a trained action recognition model to obtain the user behavior activity type.

9. The wearable urinary incontinence monitoring system based on multi-modal sensor fusion as claimed in claim 1, wherein, The user mobile terminal comprises a urine leakage amount analysis module, configured to perform the following steps: When the sensing diaper adopts a capacitive humidity sensor, obtain the urine leakage amount according to the capacitive change amount corresponding to different urine leakage amounts according to the capacitive change amount before and after each urine loss; When the sensing diaper adopts a resistive humidity sensor, obtain the urine leakage amount according to the area enclosed by the resistance change curve before and after each urine loss.

Images