Wearable core temperature monitoring device, system and method applied to sleep

Abstract

The invention discloses a wearable core temperature monitoring device, system and method applied to sleep, and relates to the field of sleep physiological parameter monitoring. The device comprises a temperature measuring sensor, an external auditory canal fixing device, a communication device, a control box, a power supply, a switch and a shell module. The sensor adopts a non-contact infrared thermopile sensor and is positioned at a position 15-25 mm away from an external auditory canal to collect the temperature of the tympanic membrane, and the precision reaches 0.01 DEG C; the fixing module is made of moldable silica gel, is matched with the ear canal and prevents blocking. The system comprises a data interaction layer, a processing layer and an application layer, and the processing layer realizes sleep stage classification and anomaly recognition through a CNN + BiLSTM + Attention model. The method comprises the steps that the wearable device collects data after self-inspection, the data is processed, displayed and early warned through the terminal, and suggestions are generated through data backup. The device is non-invasive and non-inductive to wear, can continuously and accurately monitor, and assists sleep management and health early warning.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of sleep physiological parameter monitoring, and in particular to a wearable core temperature monitoring device, system and method applied to sleep. BACKGROUND

[0002] Core temperature monitoring is of great significance in the field of medical health, especially in sleep research. Core temperature is the core parameter of body temperature regulation, accurately reflecting the temperature of internal organs, and maintaining stable internal temperature even in the case of external environmental changes. Central body temperature regulation relies on signals from core and peripheral thermal receptors, processed by the central nervous system, especially the hypothalamus. Core temperature is an important indicator for evaluating the internal physiological state of the human body, closely related to sleep quality, physiological rhythm and disease warning. In the sleep state, the human body is less affected by the outside world, which is the best period to judge the body state. Generally, heat is transferred from the core to the skin driven by blood flow, and the reduction of core temperature helps to promote the onset of sleep and enter deeper sleep stages. Abnormal core temperature changes can cause difficulty falling asleep, sleep interruption or decreased sleep quality. In addition, the circadian rhythm of core temperature is synchronized with the biological clock, and monitoring its changes during sleep helps to evaluate the stability of the physiological rhythm and prevent and improve rhythm disorders related problems. In terms of disease warning, many diseases can cause abnormal fluctuations in core temperature, such as fever in infectious diseases and hypermetabolism in metabolic diseases.

[0003] Continuous measurement of core body temperature has wide application in clinical medicine, such as surgery and nursing scenes. ISO 9886:2004 is commonly used as a reference as a general physiological standard. Common core temperature measurement methods include placing a thermometer in the core of the body, such as the rectum, esophagus, bladder cavity, pulmonary artery circulating blood and brain ventricle. Among them, rectal temperature is considered the "gold standard" for estimating core body temperature, with high accuracy and stability, but its measurement accuracy is affected by insertion depth and thermal gradient, usually requiring insertion of more than 14 centimeters into the rectum to approach the arteries near the pelvis. In addition, this measurement method can cause physiological discomfort and is not suitable for long-term monitoring. Placing a thermometer in the armpit or sublingual is also a common method for estimating core body temperature, but these sites are not truly "core body". Due to the invasiveness and interference factors of activity and environment, the core temperature measured by these methods has a large error. Sublingual temperature is easily affected by factors such as smoking or drinking hot or cold beverages before measurement, and there is a risk of cross-infection or tearing of the oral mucosa. During sleep, sublingual measurement also causes difficulty in closing the mouth, causing discomfort. Axillary temperature measurement has low accuracy, with temperature readings usually lower than the true core temperature, and is easily affected by environmental temperature, sweat and evaporation.

[0004] Currently, the existing wearable devices mostly use contact skin surface temperature monitoring, but this method is easily affected by environmental temperature and skin blood flow, and cannot accurately reflect the core temperature. In addition, the market lacks core temperature monitoring devices for sleep state, which not only need to accurately record, but also need to be able to realize automatic continuous measurement and recording in sleep; most importantly, it will not cause discomfort during wearing and affect sleep state and sleep behavior.

[0005] Traditional sleep monitoring mostly relies on polysomnography (PSG), which is complex, high-cost and not portable. In recent years, sleep staging methods based on temperature, heart rate and other physiological signals have gradually emerged, but there are still problems such as large signal noise, weak model generalization ability, and incomplete evaluation indicators.

[0006] Therefore, in view of this background, it is of great academic value and application prospect to develop a non-invasive, continuous and accurate wearable core temperature monitoring device for sleep. SUMMARY

[0007] To solve the above problems, the present application aims to provide a wearable core temperature monitoring device for sleep, which can provide real-time and comprehensive health monitoring data for users, help improve sleep quality, adjust physiological rhythms, and early warning and prevention of diseases, which is of great significance for promoting health management and disease prevention and control.

[0008] To achieve the above purpose, the technical scheme of the present application is as follows:

[0009] A wearable core temperature monitoring device for sleep, comprising a temperature measurement sensor module, an external auditory canal fixing module, a communication module, a control box module, a power module, a switch module and a shell component;

[0010] The temperature measurement sensor module is positioned at a preset distance from the auricle end to the tympanic membrane end of the external auditory canal, used for collecting tympanic membrane temperature and converting the temperature signal into an electrical signal, using a non-contact infrared sensor, with a measurement accuracy of not less than 0.1 DEG C and a measurement distance of not less than the average length of the external auditory canal, and the sensor size is adapted to the space of the external auditory canal to avoid extrusion of the ear canal;

[0011] The external auditory canal fixing module is made of flexible skin-friendly material, wrapped outside the temperature measurement sensor module, connected with the shell component and replaceable, which can be shaped according to the size and shape of the wearer's ear canal to fix the temperature measurement sensor module, and at the same time avoid blocking the sensor in the ear canal;

[0012] The communication module at least includes a short-range wireless communication submodule and a long-distance data transmission submodule, used for realizing the transmission of temperature data between the device and the terminal equipment, server;

[0013] The control box module is built-in with a master control chip and processes data through a data processing module, which includes a signal acquisition unit and a signal processing unit, respectively used for converting temperature signals and filtering, amplifying and encoding processing of signals;

[0014] The power module is a rechargeable power supply, and the single charging use duration is not less than the average duration of the sleep cycle. The power module is connected with other modules through a power line to avoid displacement of the device.

[0015] The switch module includes a general switch for controlling the start and stop of the device and a test switch for verifying data and wearing position, which is independently set from each functional module to prevent accidental touch.

[0016] The shell member is used for connecting and carrying all modules, and provides wear protection.

[0017] Further, the temperature measurement sensor module is fixed in the shell member through a temperature measurement sensor module connector and wrapped in the external ear canal fixation module for measuring the core temperature in the ear canal. The preset positioning distance of the temperature measurement sensor module is 15-25 mm. The non-contact infrared sensor is an infrared thermopile sensor, the measurement accuracy can reach 0.01℃, the measurement distance is 20-30 mm, the sensor diameter is 5.0-6.2 mm, and the length is 3.8-4.4 mm.

[0018] Further, the external ear canal fixation module is connected with the shell member through an external ear canal fixation module connector. The flexible skin-friendly material is silicone. The short-range wireless communication submodule of the communication module is a Bluetooth module, which uses 2.4-2.85GHz ISM frequency band communication. The remote data transmission submodule is a WIFI module, including but not limited to ESP8266-01S.

[0019] Further, the master control chip of the control box module is an STM32 series single-chip microcomputer, including but not limited to STM32F103C8T6. The control box module also integrates a mainboard indicator light for displaying the sensor state. The power module is a 5V lithium battery, and the single charging use duration is not less than 12 hours.

[0020] Further, the test switch of the switch module can control the device to continuously collect 3-20 times of temperature data within a preset time, and through the correction algorithm of the data processing module and the multi-factor dynamic compensation mechanism, the influence of the ear canal environment interference and the sensor self-heating on the temperature measurement is eliminated, which is used for verifying the data collection accuracy and adjusting the wearing position.

[0021] In order to achieve the above object, the application further discloses a monitoring system of a wearable core temperature monitoring device, comprising a data interaction layer, a data processing layer and a functional application layer.

[0022] The data interaction layer is used for receiving core temperature data collected and processed by the wearable core temperature monitoring device, and simultaneously realizing backup storage of the data to a server.

[0023] The data processing layer is internally provided with a deep learning prediction model, which realizes sleep stage classification, continuous value prediction and health anomaly identification by combining an attention mechanism with a multi-scale feature extraction network and a time sequence dependence capturing network.

[0024] The functional application layer comprises a data visualization module, an anomaly early warning module and an intelligent interconnection module, which are respectively used for displaying temperature changes and sleep structure, grading abnormal states and generating personalized suggestions by linking intelligent home furnishing.

[0025] Further, the deep learning prediction model of the data processing layer performs the following algorithm steps: a, normalizing preprocessing of core temperature data, scaling the data to a preset range;

[0026] b, extracting time domain features and frequency domain features, the time domain features comprising original temperature values, time information and statistical features, and the frequency domain features comprising a preset number of amplitude components of Fourier transform;

[0027] c, extracting enhanced features, including temperature change rate, moving average fluctuation and temperature gradient consistency; d, extracting data segments of a preset length by a sliding window, calculating fusion features and splicing with original temperature data;

[0028] e, learning the importance weight of each time step by using an attention mechanism, focusing on key time segments;

[0029] f, retaining a preset number of key features by a feature selection algorithm;

[0030] g, extracting local temperature patterns by using a convolutional neural network layer, and capturing time sequence dependence before and after by using a bidirectional long short-term memory network;

[0031] h, outputting sleep stage classification results and continuous value prediction results.

[0032] Further, the calculation formula of the normalization processing in step a is:

[0033]

[0034] wherein, X represents a sensor reading, X min and X max respectively represent the minimum value and the maximum value observed in the entire data set.

[0035] Further, in step c:

[0036] The temperature change rate R c The calculation formula is: Wherein, T i+1 Is the i+1th temperature; T i Is the ith temperature; n is the nth temperature sampling value;

[0037] The calculation formula of the moving average fluctuation W m

[0038] W m =std(MA3(T)), wherein MA3(T) is a three-point moving average sequence;

[0039] The calculation formula of the temperature gradient consistency Cd is:

[0040] Wherein, sign(ΔT) is the change direction sequence.

[0041] In order to achieve the above purpose, the application further discloses a monitoring method of a wearable core temperature monitoring device, comprising the following steps:

[0042] S1, wearing the device and starting the total switch of the switch module, the power module supplies power to each module, the communication module starts and searches for a matching terminal device;

[0043] S2, after the communication connection is successful, the data verification is triggered through the test switch of the switch module or the device automatically executes the self-checking program, the temperature measurement sensor module continuously collects temperature data to judge whether the wearing position is correct, and if it is deviated, it is prompted to adjust;

[0044] S3, after wearing correctly, the temperature measurement sensor module collects the tympanic membrane temperature data at a preset frequency, and transmits the data to the terminal through the communication module after being processed by the control box module;

[0045] S4, the terminal system visually displays the data, simultaneously runs the deep learning model to analyze the sleep structure and health status, and if an abnormality is identified, a graded early warning is started;

[0046] S5, after the monitoring is completed, the data is synchronously backed up to the server, and the system generates health suggestions in combination with the sleep habits of the user.

[0047] Beneficial effects: The core temperature collection position used by the application is the tympanic membrane temperature in the deep part of the external auditory canal, the tympanic membrane temperature reflects the esophageal temperature and the temperature of the hypothalamus, the branch of the carotid artery reaching the brain extends beside the tympanic membrane; in addition, the tympanic membrane can be easily accessed from the ear canal without physiological discomfort and incision, and is suitable for daily and long-term continuous monitoring;

[0048] ​The temperature measurement sensor module used in the application adopts a non-contact infrared thermocouple sensor, avoids the temperature gradient caused by the ear canal length and environmental factors of the contact sensor, and improves the measurement accuracy; at the same time, the discomfort and the influence on the sleep state and behavior caused by long-time contact are avoided;

[0049] The data processing module used in the application adopts a correction algorithm to avoid the heat accumulation caused by the ear canal environment and sensor operation, and avoid the influence of long-time monitoring on the accuracy of collected data;

[0050] The communication module and the power module used in the application can minimize the volume of the wearable device, and use a distributed monitoring, storage and power supply mode. The influence of the wearable device on the sleep state and sleep behavior is minimized, the discomfort is reduced, and a "non-sensing" sleep core temperature monitoring device is provided;

[0051] The external ear canal fixing module used in the application can effectively avoid scratching the external ear canal and foreign body sensation of the external ear canal of the temperature measurement sensor module shell, expand the external ear canal, avoid the blocking of the sensor by the ear canal contents, and shape according to the size and shape of the wearer's ear canal to better fit the auricle, fix the temperature sensor module, and avoid the receiving position of the temperature measurement sensor module from shifting;

[0052] The communication module used in the application can simultaneously store, view and post-process information in various ways such as personal mobile terminals and computers, providing more application space for the future;

[0053] The application is simple to deploy and has low cost as a wearable device, and is convenient for wide range and long time use. Users can monitor at any time and anywhere without affecting normal life and sleep. Through automatic identification and analysis of machine learning algorithms, scientific and accurate sleep evaluation and recovery suggestions are provided, and the demand and processing time of manual intervention are reduced;

[0054] The monitoring system and method used in the application are suitable for daily sleep monitoring and health management of ordinary people, help users understand their own sleep status, and improve sleep quality. It can be used in hospitals, nursing homes and other medical institutions to monitor and manage the sleep of patients or the elderly, and assist doctors in diagnosis and treatment. For professional groups (such as pilots, drivers, etc.) who need to work for a long time or in special environments, the health status such as core temperature is monitored to ensure their working state and safety.

[0055] The deep learning neural network model used in the application has the advantages of automatic feature extraction and time series modeling, can better capture complex time dependence and non-linear relationships, and reduces the need for manual feature engineering. The model adopts an end-to-end training method, and optimizes the training process through early stopping and learning rate decay strategies to ensure stability and generalization ability on large-scale sleep data sets. In addition, with the help of SHAP and other methods, the model is explained and analyzed, making the decision-making process of the model more transparent, facilitating subsequent clinical application and promotion. BRIEF DESCRIPTION OF DRAWINGS

[0056] The accompanying drawings, which form a part of the present application, are intended to provide further understanding of the present application and are incorporated herein for explanation by illustrating the specific embodiments of the present application. The description of the present application and its accompanying drawings are intended to explain the present application and should not be construed as an improper limitation of the present application. In the drawings:

[0057] Figure 1 Wearing position diagram of the wearable core temperature monitoring device for sleep according to the embodiment of the present application;

[0058] Figure 2 Structure diagram of the wearable core temperature monitoring device for sleep according to the embodiment of the present application;

[0059] Figure 3 Exploded view of the wearable core temperature monitoring device for sleep according to the embodiment of the present application;

[0060] Figure 4 Control circuit board diagram of the control box module of the wearable core temperature monitoring device for sleep according to the embodiment of the present application;

[0061] Figure 5 Main flowchart of the monitoring method of the wearable core temperature monitoring device for sleep according to the embodiment of the present application;

[0062] Figure 6 Comparison diagram of core temperature and sleep structure of the monitoring of the wearable core temperature monitoring device for sleep according to the embodiment of the present application (8 hours);

[0063] Figure 7 Comparison diagram of prediction and real sleep structure of the wearable core temperature monitoring device for sleep according to the embodiment of the present application (8 hours). DETAILED DESCRIPTION

[0064] It should be noted that the embodiments in the present application and the features in the embodiments can be combined with each other without conflict.

[0065] The present application will be described in detail below with reference to the accompanying drawings and in combination with the embodiments.

[0066] Embodiment 1

[0067] Referring to Figures 1-4 : A wearable core temperature monitoring device applied to sleep, comprising a temperature measurement sensor module 1, an external auditory canal fixing module 2, a communication module 3, a control box module 4, a power module 5, a switch module 6 and a shell component 7;

[0068] The temperature measurement sensor module 1 is positioned at a preset distance from the auricle end to the tympanic membrane end of the external auditory canal, used for collecting the tympanic membrane temperature and converting the temperature signal into an electrical signal, adopting a non-contact infrared sensor, with a measurement accuracy not less than 0.1℃ and a measurement distance not less than the average length of the external auditory canal, and the sensor size is adapted to the space of the external auditory canal to avoid extrusion of the ear canal;

[0069] The external auditory canal fixing module 2 is made of flexible skin-friendly material, wrapped on the outside of the temperature measurement sensor module 1, connected with the shell component 7 and replaceable, which can be shaped according to the size and shape of the wearer's ear canal to fix the temperature measurement sensor module 1, and at the same time avoid blocking the sensor of the ear canal;

[0070] The communication module 3 at least includes a short-range wireless communication submodule and a long-distance data transmission submodule, used for realizing the transmission of temperature data between the device and the terminal equipment, server;

[0071] The control box module 4 is internally provided with a main control chip 401 and processes data through a data processing module, and the data processing module includes a signal acquisition unit and a signal processing unit, respectively used for converting temperature signals and filtering, amplifying and encoding processing of signals;

[0072] The power module 5 is a rechargeable power supply, with a single charging use time not less than the average length of the sleep cycle, connected with other modules through a power line 501 to avoid displacement of the device;

[0073] The switch module 6 includes a master switch 601 for controlling the start and stop of the device and a test switch 602 for checking data and wearing position, which is independently set with each functional module to prevent accidental touch;

[0074] The shell component 7 is used to connect and carry all modules, and at the same time provides wear protection.

[0075] Strong measurement accuracy: non-contact infrared sensor is adopted, with a measurement accuracy not less than 0.1℃, and the measurement distance is adapted to the average length of the external auditory canal, which can accurately collect the tympanic membrane temperature to reflect the core temperature, avoiding the problem of contact sensor affected by environment and skin blood flow.

[0076] Wearing adaptability: The flexible skin-friendly material of the external auditory canal fixing module can be shaped and fixed according to the size and shape of the wearer's ear canal, avoiding squeezing the ear canal and causing discomfort, and preventing the ear canal from blocking the sensor, and adapting to the needs of different users.

[0077] Flexible data transmission: The communication module has both short-range and long-range transmission submodules, enabling data interaction between the device and the terminal and server, meeting the real-time viewing and backup storage needs.

[0078] Good endurance and stability: The power module is rechargeable, with a single endurance not less than the average length of the sleep cycle, and is connected through a power cord to avoid device displacement; the switch module is independently set to prevent accidental touch and ensure stable use.

[0079] Structural integration is reliable: The shell component uniformly carries all modules, providing wear protection, and the modules work together to continuously and stably monitor the core temperature.

[0080] In a specific example, the temperature measurement sensor module 1 is fixed in the shell component 4 through the temperature measurement sensor module connector 101 and wrapped in the external auditory canal fixing module 5, used to measure the core temperature in the ear canal, and the preset positioning distance of the temperature measurement sensor module 1 is 15-25 mm. The non-contact infrared sensor is an infrared thermopile sensor, with a measurement accuracy of 0.01°C, a measurement distance of 20-30 mm, a sensor diameter of 5.0-6.2 mm, and a length of 3.8-4.4 mm.

[0081] It should be noted that the infrared thermopile sensor is positioned at about 20 mm from the auricle end to the tympanic membrane end of the external auditory canal, used to collect the tympanic membrane temperature and convert the temperature signal into an electrical signal, collected every 30 seconds; the core temperature is determined by receiving the infrared radiation energy of the tympanic membrane through the non-contact element, with a measurement distance of 20 mm longer than the average length of the external auditory canal, avoiding errors caused by temperature gradient and measurement distance in the external auditory canal; the sensor diameter is 5.60±0.15 mm and the length is 4.10±0.15 mm, which is smaller than the average diameter of the auricle end of the external auditory canal, avoiding the compression of the ear canal during use.

[0082] In a specific example, the external auditory canal fixing module 2 is connected to the shell component 7 through the external auditory canal fixing module connector 201, and the flexible skin-friendly material is silicone. The short-range wireless communication submodule of the communication module 3 is a Bluetooth module, which communicates using the 2.4-2.85 GHz ISM frequency band, and the long-range data transmission submodule is a WIFI module, including but not limited to ESP8266-01S.

[0083] It should be noted that the external auditory canal fixing module is a shapeable skin-friendly silicone material with certain supportability, which is connected with the shell member and can be replaced; the external auditory canal fixing module is wrapped outside the temperature measurement sensor module to avoid scratching the external auditory canal and foreign body sensation of the external auditory canal, expand the external auditory canal, and avoid the occlusion of the sensor by the contents in the ear canal; the shape can be shaped according to the size and shape of the wearer's ear canal to better fit the auricle, fix the temperature sensor module, and avoid the displacement of the temperature measurement sensor module receiving position.

[0084] The Bluetooth module is used for short-distance communication for data transmission of the measurement site, uses short wave high frequency (UHF) radio waves, and communicates through the ISM frequency band of 2.4-2.85 GHz; the WIFI module is connected to the main control chip of the main control module, and the model is ESP8266-01S; after connecting to the local network, information is received and sent to the server end, the received message is fed back to the main control chip for corresponding operation, and the sent message is transmitted to the server end for storage; the Bluetooth module as near field communication can transmit messages between the control box module and the computer or personal intelligent terminal.

[0085] In a specific example, the main control chip 401 of the control box module 4 is an STM32 series single-chip microcomputer, and the model includes but is not limited to STM32F103C8T6; the control box module 4 also integrates a mainboard indicator light 402 for displaying the sensor state; the power module 5 is a 5V lithium battery, and the single-charge use time is not less than 12 hours.

[0086] It should be noted that the control box module of the embodiment is an STM32 series single-chip microcontroller, which is internally provided with a data processing module, and the data processing module includes: a signal acquisition unit and a signal processing unit; the chip model is STM32F103C8T6; the signal acquisition unit is used to process the data output signal collected by the temperature measurement sensor module and convert it into a digital signal; the signal processing unit is used to filter, amplify, and encode the digital signal to generate a data set;

[0087] Power module: a large-capacity rechargeable lithium battery, usually a 5V lithium battery. It can power other modules, and the use time is not less than 12 hours, and it can provide power for various sleep states; it is connected with other modules through a long enough power line to avoid affecting the sleep state and the displacement and falling of the wearable device caused by sleep behavior.

[0088] In a specific example, the test switch 602 of the switch module 6 can control the device to continuously collect 3-20 times of temperature data within a preset time, and through the correction algorithm of the data processing module and the multi-factor dynamic compensation mechanism, the influence of the ear canal environment interference and the sensor self-heating on the temperature measurement is eliminated, which is used to verify the data collection accuracy and adjust the wearing position.

[0089] The test switch of the embodiment can control the wearable device to immediately continuously collect 10 times of data, which is used to verify the accuracy of data collection and the wearing position of the wearable device.

[0090] In a specific implementation, the correction algorithm includes the following steps:

[0091] Baseline calibration and environmental parameter modeling:

[0092] The algorithm first establishes an initial temperature baseline by triggering 10 times of continuous sampling through the test switch during the device initialization stage (such as the first wearing or the first use of each day). At the same time, based on the physiological structure characteristics (such as length, average temperature gradient) of the external ear canal and the characteristics (such as measurement distance and temperature attenuation coefficient) of the infrared sensor, an environmental interference model is pre-constructed to record the temperature deviation law under different ear canal positions and air flow states.

[0093] Real-time interference factor monitoring and compensation:

[0094] During the continuous monitoring process, the algorithm dynamically corrects in the following ways:

[0095] For ear canal environmental interference: through the temperature fluctuation frequency and amplitude collected by the sensor, the ear canal air flow caused by head posture change (manifested as short-term temperature sudden change) is identified, and a preset gradient compensation coefficient is combined to linearly correct the temperature value deviating from the baseline;

[0096] For sensor heating effect: real-time monitoring of sensor working current, continuous running time and other parameters, according to the pre-labeled “working time-temperature offset” curve, the measurement value offset caused by the sensor self-heating is reversely compensated (such as automatically subtracting a basic offset of 0.02℃ for every 1 hour of continuous working).

[0097] Data fusion and smoothing processing:

[0098] The algorithm performs sliding window (window size is 5 sampling points, i.e. 2.5 minutes) smoothing processing on the compensated temperature data to eliminate high-frequency noise; at the same time, in combination with the historical data trend (such as the temperature change slope in the last 30 minutes), the abnormal jump value (exceeding the normal physiological fluctuation range ±0.3℃) is secondarily verified, and finally the stable core temperature value is output, which provides reliable input for the subsequent sleep analysis model.

[0099] The mechanism realizes dynamic adaptation to complex ear canal environment and device self characteristics by combining hardware parameter monitoring with software algorithm, and ensures that the accuracy of core temperature data in long time sleep monitoring is maintained within ±0.05℃.

[0100] Embodiment 2

[0101] To achieve the above purpose, the embodiment also discloses a monitoring system of a wearable core temperature monitoring device, which comprises a data interaction layer, a data processing layer and a functional application layer.

[0102] The data interaction layer is used for receiving core temperature data collected and processed by the wearable core temperature monitoring device, and simultaneously realizing backup storage of the data to a server.

[0103] The data processing layer is internally provided with a deep learning prediction model, which realizes sleep stage classification, continuous value prediction and health abnormality identification by combining an attention mechanism with a multi-scale feature extraction network and a time sequence dependence capturing network.

[0104] The functional application layer comprises a data visualization module, an abnormality early warning module and an intelligent interconnection module, which are respectively used for displaying temperature change and sleep structure, classifying and warning abnormal states, and generating personalized suggestions in linkage with smart home.

[0105] The deep learning prediction model of the embodiment is specifically a CNN+BiLSTM+Attention neural network deep learning prediction model, which realizes sleep stage classification and continuous value prediction by combining an attention mechanism with a multi-scale convolutional neural network and a bidirectional LSTM, and provides multi-index evaluation and visual analysis functions for automatically identifying sleep cycles and abnormal data; the identified sleep cycles are marked with different color or shape blocks, such as light sleep, deep sleep, rapid eye movement (REM) sleep, non-rapid eye movement (NREM) and wakefulness, and the sleep structure is clearly displayed; the core temperature data are used for identifying and warning possible health abnormality states in sleep.

[0106] In a specific example, the deep learning prediction model of the data processing layer performs the following algorithm steps:

[0107] a. Normalizing the core temperature data for preprocessing, and scaling the data to a preset range;

[0108] b. Extracting time domain features and frequency domain features, wherein the time domain features comprise original temperature values, time information and statistical features, and the frequency domain features comprise a preset number of amplitude components of Fourier transform;

[0109] c. Extracting enhanced features, including temperature change rate, moving average fluctuation and temperature gradient consistency; d. Extracting data segments of a preset time length through a sliding window, calculating fusion features and splicing with original temperature data;

[0110] e. Learning the importance weight of each time step using attention mechanism, focusing on key time segments;

[0111] f. Retaining a preset number of key features through feature selection algorithm;

[0112] g. Extracting local temperature patterns using convolutional neural network layers, capturing temporal dependencies before and after using bidirectional long short-term memory network;

[0113] h. Outputting sleep stage classification results and continuous value prediction results.

[0114] It should be noted that the deep learning prediction model of the embodiment is specifically a CNN+BiLSTM+Attention neural network deep learning prediction model, and a standard training set of the CNN+BiLSTM+Attention neural network deep learning prediction model needs to be constructed, which includes time labels, whole night sleep structure and conversion data, and corresponding core temperature changes in the whole night sleep; the constructed standard training set is preprocessed, including but not limited to: denoising, screening, division, dimensionality reduction and normalization; the CNN+BiLSTM+Attention neural network deep learning prediction model is trained based on the standard training set.

[0115] The input data of the deep learning prediction model of the embodiment is derived from the continuous temperature signal collected by the non-invasive tympanic membrane temperature monitoring device, and the data is organized at an interval of 30 seconds. The input of each time step is composed of 1 original temperature data and 12 fusion features, totaling 13 dimensions. The temperature time series data is processed by a sliding window, each window contains 30 consecutive sampling points, and the feature values in the time domain and frequency domain are calculated in turn, including: original temperature value, time information (absolute time, time after awakening, sleep latency), statistical features (mean, standard deviation, extreme value, difference index) and the first four amplitude components of Fourier transform.

[0116] Prediction optimization of WAKE stage: the prediction of WAKE stage is unstable mainly because as the sleep deepens, the core temperature gradually decreases, and when a short wake-up occurs during sleep, the core temperature increases limitedly, which is easy to overlap with the core temperature interval of other stages; at the same time, the wake-up time during sleep is usually short, and the time series index and sliding window may ignore these short states, causing the model to misjudge. To further improve the prediction effect, the model introduces time information (absolute time, time after awakening, sleep latency) in the feature extraction stage to more accurately label the wake-up and NREM period.

[0117] It should be noted that the WAKE stage is a wake state, i.e., a non-sleep stage during sleep or before sleep, and the NREM stage refers to a non-rapid eye movement sleep stage, which includes three stages of sleep: N1 stage, i.e., a light sleep stage, a stage for transitioning from wake to sleep; N2 stage, i.e., a moderate sleep stage, a main part of sleep; and N3 stage, i.e., a deep sleep stage, a stage with the most significant recovery effect.

[0118] In a specific example, the calculation formula of the normalization processing in step a is:

[0119]

[0120] wherein X represents a sensor reading, X min and X max respectively represent the minimum value and the maximum value observed in the entire data set.

[0121] In a specific example, the calculation formula of the temperature change rate R c in step c is:

[0122] i+1 wherein T i is the i+1th temperature; T m is the ith temperature; and n is the nth temperature sampling value.

[0123] The calculation formula of the moving average fluctuation W m in step d is:

[0124] W d =std(MA3(T)), wherein MA3(T) is a three-point moving average sequence.

[0125] The calculation formula of the temperature gradient consistency C -4 in step e is:

[0126] wherein sign(ΔT) is a change direction sequence.

[0127] The extracted basic features and enhanced features are arranged and spliced along the time sequence with the original temperature data to form a feature matrix;

[0128] It should be noted that multi-branch one-dimensional convolution (kernel size of 3, 5, and 7) is used to extract multi-scale features, which are input into a time attention layer after being connected in series for weight optimization, and then processed by separable convolution and pooling (including Dropout). Subsequently, a double-layer BiLSTM (384 and 128 units) is used to capture long-time dependence, and finally two task heads are output through a shared fully connected layer: a softmax classification result of 5 sleep stages and a regression prediction of a continuous value. The network uses an Adam optimizer (learning rate 1e​-4 ), the loss function is weighted cross-entropy (0.8) and mean square error (0.05), and L1 regularization and Dropout are introduced to suppress overfitting.

[0129] The model accuracy is verified by multi-dimensional indicators. The data distribution is reasonably ensured by using a stratified random division strategy. The accuracy improvement mechanism includes three levels of optimization: at the feature level, Z-score standardization is used to eliminate dimensional differences, frequency-time domain complementary features and stage transition constraints are introduced; at the model level, bidirectional LSTM is used to capture bidirectional time sequence dependence, attention mechanism is used to focus on key frames for decision-making, and Dropout layer is set to suppress overfitting; at the training level, mixed precision acceleration, adaptive learning rate decay and class weight adjustment strategies are implemented. The stability and generalization ability on large-scale sleep data sets are guaranteed. In addition, SHAP and other methods are used to explain and analyze the model, making the decision-making process of the model more transparent, facilitating subsequent application and promotion.

[0130] Embodiment 3

[0131] To achieve the above purpose, see Figure 5 The embodiment also discloses a monitoring method of the wearable core temperature monitoring device, comprising the following steps:

[0132] S1, wearing the device and starting the general switch 601 of the switch module 6, the power module 5 supplies power to each module, the communication module 3 is started and searches for a matching terminal device;

[0133] S2, after successful communication connection, triggering data verification or device automatic self-checking program through the test switch 602 of the switch module 6, the temperature measurement sensor module 1 continuously collects temperature data to judge whether the wearing position is correct, and if it is offset, it is prompted to adjust;

[0134] S3, after correct wearing, the temperature measurement sensor module 1 collects the tympanic membrane temperature data at a preset frequency, and transmits the data to the terminal through the communication module 3 after being processed by the control box module 4;

[0135] S4, the terminal system visually displays the data, and simultaneously runs a deep learning model to analyze sleep structure and health status, and if an abnormality is identified, a graded early warning is started;

[0136] S5, after the monitoring is completed, the data is synchronously backed up to a server, and the system generates health suggestions in combination with the user's sleep habits.

[0137] In the specific implementation, as shown in Figure 5 In the sleep preparation stage, the user can, for example Figure 1In the illustrated manner, the wearable core temperature detection device is worn on the ear, and the power switch is turned on to boot up; the communication module starts communication broadcasting to search for terminals using the matching software until a successful connection is made.

[0138] After a successful connection, the device automatically runs a self-checking program to continuously record 10 core temperature data, and detects whether the data is within the correct range; if the data is abnormal and caused by wearing offset, the terminal software sends a pop-up window warning and guides the user to adjust the wearing position; if the self-checking is timed out or the position is adjusted again, the detection switch can be pressed to manually correct; if the software pop-up window prompts the device to be damaged due to connection, hardware error, etc.

[0139] After wearing correctly, the monitoring can be manually started in the software or after the sleep starts, the software automatically identifies the sleep cycle and starts monitoring; after the data is preprocessed by the control box module, it is uploaded to the software for identification and post-processing; at the same time, the data will be uploaded to the cloud storage backup at the same time, and the data optimization adjustment system will be run to help build user sleep preferences and personalized identification models;

[0140] In the terminal software, the sleep core temperature information and sleep structure information can be viewed in real time on multiple terminals; in addition, abnormal states are identified, and users or auxiliary personnel are reminded to intervene through pop-up windows, sound alarms, etc.; according to the severity of the abnormality, the warning is divided into different levels, such as mild abnormality, moderate abnormality and severe abnormality, and corresponding warning measures are taken;

[0141] The software reserves a multi-modal interface, which can interconnect with other smart home devices (such as smart beds, etc.) to adjust the sleep environment and interface environment; combined with the user's health status, living habits and specific reasons for abnormal conditions, personalized recovery suggestions are generated, such as adjusting the sleep environment (temperature, humidity, light, etc.), improving sleep habits (such as relaxation before sleep, regular work and rest, etc.), and appropriate exercise; when necessary, the system can also provide remote consultation or advice from professional doctors to help users better solve sleep problems.

[0142] In a specific implementation, the core temperature and sleep structure data are measured simultaneously for the whole night (8 hours) under the working conditions of a standard artificial climate chamber (temperature 25±5℃, humidity 55±5%). The sleep structure is divided by recording electroencephalogram signals, eye movement signals, etc. through the PSG system. Figure 6 The comparison chart of the whole night (8 hours) core temperature and sleep structure shows that the core temperature monitored by the wearable core temperature monitoring device changes synchronously with the sleep structure throughout the night, and continuously decreases as the sleep deepens.

[0143] For example, Figure 7For the whole night (8 hours) prediction comparison with the real sleep structure, the contrast chart shows the prediction effect of the deep learning model used by the monitoring system of the wearable core temperature monitoring device. The prediction accuracy (Accuracy) reaches 0.839, the precision (Precision) reaches 0.749, the recall (Recall) reaches 0.727, and the F1 score reaches 0.717; the regression consistency reaches 0.7612. The kappa coefficient reaches 0.756; it is proved that the model of the embodiment can achieve excellent prediction performance in both classification and regression tasks, and the prediction performance can be further strengthened through continuous learning and personalized parameter input.

[0144] It should also be noted that the electronic device also includes a terminal device, which can also be referred to as a terminal, user equipment (UE), mobile station (MS), mobile terminal (MT), etc. The terminal device can be a mobile phone, a smart television, a wearable device, a tablet computer (Pad), a computer with wireless transceiver function, a virtual reality (VR) terminal device, an augmented reality (AR) terminal device, a wireless terminal in industrial control, a wireless terminal in self-driving, a wireless terminal in remote medical surgery, a wireless terminal in smart grid, a wireless terminal in transportation safety, a wireless terminal in smart city, a wireless terminal in smart home, etc. The embodiments of the present application do not limit the specific technology and specific device form of the terminal device.

[0145] The above only describes the preferred embodiments of the present application and is not intended to limit the present application. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application shall be included in the protection scope of the present application.

Claims

1. A wearable core temperature monitoring device for use in sleep, characterized in that, The ear temperature measurement device comprises a temperature measurement sensor module (1), an external ear canal fixing module (2), a communication module (3), a control box module (4), a power module (5), a switch module (6) and a shell component (7). The temperature measurement sensor module (1) is positioned at a preset distance from the auricle end to the eardrum end of the external ear canal, used for collecting the eardrum temperature and converting the temperature signal into an electric signal, adopting a non-contact infrared sensor, with a measurement accuracy of not less than 0.1℃ and a measurement distance of not less than the average length of the external ear canal, and the sensor size is adapted to the space of the external ear canal to avoid extrusion of the ear canal. The external ear canal fixing module (2) is made of flexible skin-friendly material, wrapped outside the temperature measurement sensor module (1), connected with the shell component (7) and replaceable, and can be shaped according to the size and shape of the wearer's ear canal to fix the temperature measurement sensor module (1), while avoiding blocking the sensor of the ear canal. The communication module (3) at least includes a short-range wireless communication submodule and a remote data transmission submodule, used for realizing the transmission of temperature data between the device and the terminal equipment and the server. The control box module (4) is internally provided with a main control chip (401) and processes data through a data processing module, which includes a signal acquisition unit and a signal processing unit, respectively used for converting temperature signals and filtering, amplifying and encoding processing of signals. The power module (5) is a rechargeable power supply, with a single charging use time of not less than the average length of the sleep cycle, connected with other modules through a power line (501) to avoid displacement of the device. The switch module (6) includes a master switch (601) for controlling the start and stop of the device and a test switch (602) for checking data and wearing position, independently arranged with each functional module to prevent false touch. The shell component (7) is used for connecting and carrying all modules, and providing wear protection.

2. The wearable core temperature monitoring device of claim 1, wherein, The temperature measurement sensor module (1) is fixed in the shell component (4) through a temperature measurement sensor module connector (101) and wrapped in the external ear canal fixing module (5), used for measuring the core temperature in the ear canal; the preset positioning distance of the temperature measurement sensor module (1) is 15-25mm, the non-contact infrared sensor is an infrared thermocouple sensor, the measurement accuracy can reach 0.01℃, the measurement distance is 20-30mm, the sensor diameter is 5.0-6.2mm, and the length is 3.8-4.4mm.

3. The wearable core temperature monitoring device of claim 1, wherein, The external ear canal fixing module (2) is connected with the shell component (7) through an external ear canal fixing module connector (201), the flexible skin-friendly material is silica gel, the short-range wireless communication submodule of the communication module (3) is a Bluetooth module, which communicates at 2.4-2.85GHz ISM frequency band, and the remote data transmission submodule is a WIFI module, including but not limited to ESP8266-01S.

4. The wearable core temperature monitoring device of claim 1, wherein, The main control chip (401) of the control box module (4) is an STM32 series single-chip microcomputer, including but not limited to STM32F103C8T6, and the control box module (4) is also integrated with a mainboard indicator light (402) for displaying the sensor state. The power module (5) is a 5V lithium battery, and the single-charge use time is not less than 12 hours.

5. The wearable core temperature monitoring device of claim 1, wherein, The test switch (602) of the switch module (6) can control the device to continuously collect 3-20 times of temperature data within a preset time, and through the correction algorithm of the data processing module and the multi-factor dynamic compensation mechanism, the influence of the ear canal environment interference and the sensor self-heating on the temperature measurement is eliminated, which is used to verify the data collection accuracy and adjust the wearing position.

6. A monitoring system based on the wearable core temperature monitoring device according to any one of claims 1-5, characterized in that, It comprises a data interaction layer, a data processing layer and a functional application layer. The data interaction layer is used for receiving the core temperature data collected and processed by the wearable core temperature monitoring device, and simultaneously realizing the backup storage of the data to the server. The data processing layer is internally provided with a deep learning prediction model, which realizes sleep stage classification, continuous value prediction and health anomaly identification through the combination of a multi-scale feature extraction network and a time sequence dependence capture network with an attention mechanism. The functional application layer comprises a data visualization module, an abnormal early warning module and an intelligent interconnection module, which are respectively used for displaying temperature changes and sleep structure, grading abnormal state and generating personalized suggestions for intelligent home linkage.

7. The monitoring system of claim 6, wherein, The deep learning prediction model of the data processing layer performs the following algorithm steps: a. Normalizing the core temperature data for preprocessing, scaling the data to a preset range; b. Extracting time domain features and frequency domain features, the time domain features including original temperature values, time information and statistical features, and the frequency domain features including a preset number of amplitude components of Fourier transform; c. Extracting enhanced features, including temperature change rate, moving average fluctuation and temperature gradient consistency; d. Extracting data segments of a preset length through a sliding window, calculating fusion features and splicing with original temperature data; e. Learning the importance weight of each time step using an attention mechanism, focusing on key time segments; f. Retaining a preset number of key features through a feature selection algorithm; g. Extracting local temperature patterns using a convolutional neural network layer, and capturing forward and backward time sequence dependence through a bidirectional long short-term memory network; h. Outputting sleep stage classification results and continuous value prediction results.

8. The monitoring system of claim 6, wherein, The calculation formula of the normalization processing in step a is: where X represents the sensor reading, X min and X max represent the minimum and maximum values observed across the dataset, respectively.

9. The monitoring system of claim 7, wherein, In step c, the following steps are included: Temperature change rate R c The calculation formula is: Wherein, T i+1 is the i+1th temperature; T i is the ith temperature; n is the nth temperature sampling value; Moving average volatility W m The formula for calculating is: W m = std(MA3(T)), where MA3(T) is a three-point moving average sequence; Temperature gradient consistency C d The calculation formula is: where sign(ΔT) is the sequence of change direction.

10. A monitoring method based on the wearable core temperature monitoring device according to any one of claims 1-5, characterized in that, S1, wearing the device and starting the main switch (601) of the switch module (6), the power module (5) supplies power to each module, and the communication module (3) starts and searches for a matching terminal device; S2, after successful communication connection, triggering data verification through the test switch (602) of the switch module (6) or automatically executing a self-checking program by the device, the temperature measurement sensor module (1) continuously collects temperature data to determine whether the wearing position is correct, and if not, prompting adjustment; S3, after correct wearing, the temperature measurement sensor module (1) collects tympanic membrane temperature data at a preset frequency, which is processed by the control box module (4) and transmitted to the terminal through the communication module (3). ​ S4, the terminal system visualizes the data and runs a deep learning model to analyze the sleep structure and health status. If an abnormality is identified, a graded warning is initiated. S5, after the monitoring ends, the data is synchronized and backed up to the server. The system generates health recommendations based on the user's sleep habits.

Images