Wearing state recognition method and device and computer readable storage medium

Abstract

The invention relates to the technical field of wearable devices, in particular to a wearable state recognition method and device and a computer readable storage medium, and the method comprises the following steps: obtaining a wearable device set paired with a mobile phone specified by a user, and obtaining sensor collection data through the wearable devices, determining a wearing state according to the data collected by the sensor and generating a worn device set. According to real-time physiological data and motion data acquired by the worn devices, normal wearable devices and abnormal wearable devices are recognized, the abnormal wearable devices in the worn device set are removed, and a notification receiving device set is formed. And setting priorities for the normal wearable devices, dividing the notification receiving device set into a notification device set and a silent device set according to the priorities, and finally, only one normal wearable device performs notification reminding, and the other normal wearable devices are kept silent. And when the specified mobile phone receives a new message, the steps are executed to determine the notification mode, so that interference caused by simultaneous reminding of multiple devices is reduced, and the use experience of a user is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of wearable devices, and particularly to a method and device for recognizing a wearing state and a computer-readable storage medium. Background Art

[0002] With the popularization of wearable devices, more and more people wear devices such as smart watches, smart bracelets, smart earphones, and smart rings to monitor their health conditions in real time, receive notification reminders, and improve the convenience of life. These devices are usually connected to a smart phone to receive notification reminders from the phone. A user may wear multiple devices in different scenarios, and the wearing states of these devices have a direct impact on the delivery effect of notifications. If the wearing state of a device cannot be accurately recognized, notifications may be pushed to a device that is not worn or is worn improperly, resulting in the user missing important information or being disturbed unnecessarily.

[0003] Existing solutions include setting priorities and filtering rules for notifications between devices, allowing users to customize which devices receive specific notifications. In addition, through a device preference selection mechanism, important notifications can be concentrated and sent to the device that the user uses most frequently to reduce the interference of simultaneous reminders from multiple devices. Existing solutions have some drawbacks and limitations. For example, users need to manually set the priorities and filtering rules for notifications, which is a cumbersome process, especially when the user has multiple devices and needs to configure them one by one. In addition, if a device is not worn or has insufficient power, important notifications may be sent to that device, resulting in the user missing key information. Moreover, the preference selection mechanism between devices may sometimes fail to accurately recognize the current needs of the user, especially in a multitasking or dynamic environment and may not be able to adapt to the actual usage of the user in a timely manner. There is an urgent need for a method for recognizing the wearing states of multiple devices of a user to reduce the interference of simultaneous reminders from multiple devices and improve the user experience.

[0004] Therefore, a method and device for recognizing a wearing state and a computer-readable storage medium are proposed. Summary of the Invention

[0005] The object of the present invention is to provide a wearing state recognition method, device and computer-readable storage medium. The present invention mainly starts from the following aspects: First, obtain the real-time notification status of the user-specified mobile phone. If the real-time notification status is "yes", execute the recognition method, including the following steps: Obtain the set of wearable devices currently paired with the mobile phone. Determine the wearing state of the devices through sensor data, screen out the worn devices, and obtain their corresponding real-time physiological data and motion data. Then, respectively identify normal and abnormal wearing devices based on these data, and remove the abnormal wearing devices from the set of worn devices composed of all worn devices to form a notification receiving device set. Set priorities for the normally worn devices, and divide the devices into a notification device set and a silent device set according to the priorities to ensure that only one normally worn device is used for notification reminders. In addition, a multi-modal classification model is proposed. The input layer receives physiological features, motion features and cross-modal features respectively, and uses an independent feature processing layer and a self-attention mechanism layer to deeply extract and weight process various features. The feature fusion layer uniformly fuses the weighted features, and finally performs a non-linear transformation through the classification layer and generates an abnormal probability. The expression ability and accuracy of the model in processing complex data are improved, the recognition accuracy of abnormal wearing states is effectively improved, and the reliability of the method is enhanced. Intelligently sort the normally worn devices through priority scoring, assign the device with the highest priority as the notification device, and set the remaining devices to the silent state. The priority scoring comprehensively considers the specified mode, historical usage frequency, battery power, functional relevance, user preferences, signal strength and current state of the device to ensure that notifications are only sent through the device with the highest priority score, avoiding interference caused by simultaneous reminders of multiple devices, thereby improving the user experience and device efficiency. Finally, record the historical environmental data and user response time after the notification reminder, and dynamically adjust the message notification method according to the similarity between the real-time environment and the historical data. Intelligently select the optimal notification method in different environments, and even if the notification method is not specified in advance, the response speed of the user can be maximized, further improving the user experience.

[0006] To achieve the above object, the present invention provides the following technical solutions:

[0007] According to a first aspect of the present disclosure, there is provided a wearing state recognition method, including:

[0008] Obtain the real-time notification status of the user-specified mobile phone to determine whether a new message is received;

[0009] If the real-time notification status is "yes", execute the following steps, specifically including:

[0010] Step 1: Obtain the set of wearable devices currently paired with the specified mobile phone, where the set of wearable devices includes N wearable devices with physiological feature recording functions, motion recording functions, environmental data recording functions and mobile phone synchronous notification functions;

[0011] Step 2: Obtain the sensor acquisition data through each of the wearable devices; if the sensor acquisition data meets the specified conditions, determine that the corresponding wearable device is in a worn state; obtain all the wearable devices in the worn state to get the worn devices; all the worn devices form a set of worn devices; for each of the worn devices, obtain the corresponding real-time physiological data and real-time motion data to get the first physiological data and the first motion data;

[0012] Step 3: Identify the normally worn devices and abnormally worn devices respectively according to the first physiological data and the first motion data; remove the abnormally worn devices from the set of worn devices to get the set of notification receiving devices;

[0013] Step 4: Set priorities for the normally worn devices;

[0014] Step 5: Divide the set of notification receiving devices into a set of notification devices and a set of silent devices according to the priorities; the set of notification devices includes only one of the normally worn devices and is used for notification reminders; the set of silent devices includes at least one of the normally worn devices and is used to maintain a silent state.

[0015] Further, obtaining the real-time notification status of the specified mobile phone includes:

[0016] Continuously monitor the notification center of the specified mobile phone; when it is detected that the specified mobile phone receives a new message, extract the basic information of the new message, including the message type, content summary, source application, reception timestamp, and read status; screen the basic information according to the preset data synchronization rules, determine that the new message whose basic information conforms to the data synchronization rules has a new message received, and update the real-time notification status to yes.

[0017] Further, obtaining the sensor acquisition data of the user through each of the wearable devices; if the sensor acquisition data meets the specified conditions, determining that the corresponding wearable device is in the worn state specifically is:

[0018] The sensor acquisition data includes a detected resistance value, a detected temperature value, and a detected pressure value;

[0019] For each of the wearable devices, the resistance value at the data acquisition location of the wearable device is detected by a skin resistance sensor to obtain the detected resistance value; if the detected resistance value is within a preset resistance range, a skin resistance test is performed; the temperature at the contact point between each wearable device and the skin is measured by a temperature sensor to obtain the detected temperature value; if the detected temperature value is within a preset temperature range, a temperature test is performed; the pressure value at the data acquisition location of each wearable device is detected by a pressure sensor to obtain the detected pressure value; if the detected pressure value is within a preset pressure range, a pressure test is performed;

[0020] The wearable devices that pass the skin resistance test, the temperature test, and the pressure test simultaneously are obtained to get the correctly worn devices; it is determined that the correctly worn devices are in the worn state.

[0021] Further, identifying the normally worn devices and the abnormally worn devices based on the first physiological data and the first motion data includes:

[0022] Obtain the historical normal physiological data and historical normal motion data corresponding to each wearable device in the wearable device set; the historical normal physiological data are various physiological parameters recorded by the user when the wearable device is worn normally; the historical normal motion data are various motion parameters recorded by the user when the wearable device is worn normally;

[0023] Use historical abnormal records to obtain the physiological data and motion data generated by the user when the wearable device is worn abnormally, to get historical abnormal physiological data and historical abnormal motion data;

[0024] Label the historical normal physiological data and the historical normal motion data as positive samples to get positive sample labels; label the historical abnormal physiological data and the historical abnormal motion data as negative samples to get negative sample labels; integrate the positive and negative sample data into a labeled historical physiological data set and a labeled historical motion data set respectively;

[0025] Perform timestamp synchronization and standardization processing on the labeled historical physiological data and the labeled historical motion data, and remove noise, outliers, and fill in missing values by cleaning the data, to get preprocessed historical physiological data and preprocessed historical motion data;

[0026] Extract features from the preprocessed historical physiological data and the preprocessed historical motion data to obtain physiological features, motion features, and cross-modal features; the physiological features include skin conductivity; the motion features include step frequency, step amplitude, step speed, acceleration change rate, motion direction change, and micro-motion features; the cross-modal features include heart rate-step frequency ratio and acceleration-skin conductivity ratio;

[0027] Using the physiological features, the motion features, and the cross-modal features as input data, and using the positive sample labels and the negative sample labels as training labels, train a multi-modal classification model;

[0028] Input the first physiological data and the first motion data into the trained multi-modal classification model to obtain an anomaly probability;

[0029] If the anomaly probability exceeds a preset threshold, identify the wearable device corresponding to the first physiological data and the first motion data as the abnormal wearable device;

[0030] If the anomaly probability does not exceed the preset threshold, identify the wearable device corresponding to the first physiological data and the first motion data as the normal wearable device.

[0031] Furthermore, the multi-modal classification model includes:

[0032] An input layer for receiving input data including the physiological features, the motion features, and the cross-modal features; the input layer includes three sub-input modules for receiving and processing the physiological features, the motion features, and the cross-modal features respectively;

[0033] A feature processing layer including an independent feature processing layer and a self-attention mechanism layer. The independent feature processing layer consists of three fully connected layers, each layer containing 128, 64, and 32 neurons respectively, and using the ReLU activation function. Dropout is applied between each layer to deeply extract the physiological features, the motion features, and the cross-modal features to obtain physiological depth features, motion depth features, and cross-modal depth features; the self-attention mechanism layer is used to weight the physiological depth features, the motion depth features, and the cross-modal depth features to obtain weighted features;

[0034] A feature fusion layer is used to fuse the weighted features to generate a unified feature vector;

[0035] A classification layer includes three fully connected layers for gradually compressing the feature vector generated by the feature fusion layer. Each fully connected layer contains 32, 16, and 1 neuron respectively, and performs a non-linear transformation through the ReLU activation function; the last fully connected layer of the classification layer uses the Sigmoid activation function to generate the anomaly probability.

[0036] Furthermore, classifying the notification receiving device set into the notification device set and the silent device set according to the priority specifically is:

[0037] If the normal wearable device in the receiving device set is in the specified mode, set the priority score of the normal wearable device in the receiving device set in the specified mode to 0; the specified mode includes a silent mode and a do-not-disturb mode;

[0038] If the normal wearable device in the receiving device set is not in the specified mode, obtain the historical device usage frequency, the current battery level, the device function relevance index, the user preference index, the signal strength score, and the device current state of the normal wearable device; the device current state includes an active state and an inactive state; calculate the priority score according to the historical device usage frequency, the current battery level, the device function relevance index, the user preference index, the signal strength score, and the device current state:

[0039] Sort all the normal wearable devices in the receiving device set according to the priority score, and select the normal wearable device with the highest priority score as the highest-priority device; allocate the highest-priority device to the notification device set, and allocate the normal wearable devices in the receiving device set other than the highest-priority device to the silent device set.

[0040] Further, dividing the receiving device set into the notification device set and the silent device set according to the priority further includes: if the number of the highest-priority devices is greater than one, screen the highest-priority devices based on a preset rule to determine the final notification device, and allocate the final notification device to the notification device set.

[0041] Further, the method further includes: after the worn device in the notification device set performs the notification reminder, record the historical environmental data and the user response time; the historical environmental data is the environmental data collected at the first time corresponding to the notification reminder, including the noise level and the light intensity; the response time is the time interval from the first time to the time when the user feedback operation is detected; the user feedback operation includes a specified interaction operation through the device in the notification device set or the specified mobile phone;

[0042] When the message notification method is not specified in advance, adjust the message notification method according to the historical environmental data and the user response time, including:

[0043] When there is a new message received by the specified mobile phone, the wearable device in the notification device set acquires environmental data of the current environment to obtain real-time environmental data. By calculating the similarity scores between the real-time environmental data and each piece of historical environmental data, the historical environmental data with the highest similarity score is identified to obtain the first historical environmental data; the user response time corresponding to the first historical environmental data is acquired, and the historical message notification method with the shortest user response time is selected as the message notification method.

[0044] According to a second aspect of the present disclosure, there is provided a wearable state recognition device, including:

[0045] A notification state acquisition module, configured to acquire the real-time notification state of the user's specified mobile phone to determine whether a new message is received;

[0046] A wearable device set acquisition module, configured to acquire the set of wearable devices currently paired with the specified mobile phone, where the set of wearable devices includes N wearable devices with physiological feature recording functions, motion recording functions, environmental data recording functions, and mobile phone synchronous notification functions;

[0047] A data acquisition module, configured to acquire sensor acquisition data through each wearable device; if the sensor acquisition data meets the specified conditions, it is determined that the corresponding wearable device is in a worn state; all the wearable devices in the worn state are acquired to obtain the worn devices; all the worn devices form a set of worn devices; for each worn device, the corresponding real-time physiological data and real-time motion data are acquired to obtain the first physiological data and the first motion data;

[0048] An abnormal device identification module, configured to identify normal wearable devices and abnormal wearable devices respectively according to the first physiological data and the first motion data; remove the abnormal wearable devices from the set of worn devices to obtain a set of notification receiving devices;

[0049] A priority setting module, configured to set priorities for the normal wearable devices;

[0050] A device allocation module, configured to divide the set of notification receiving devices into a notification device set and a silent device set according to the priorities; the notification device set includes only one normal wearable device for notification reminder; the silent device set includes at least one normal wearable device for maintaining a silent state;

[0051] A control module, configured to, when the notification state acquisition module detects that the real-time notification state is yes, control the device to perform the operations of the wearable device set acquisition module, the data acquisition module, the abnormal device identification module, the priority setting module, and the device allocation module.

[0052] According to a third aspect of the present disclosure, a computer-readable storage medium stores a computer program thereon, and when the computer program is executed by a processor, the steps of the method described in the first aspect of the present disclosure are implemented.

[0053] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0054] The present invention proposes a multi-modal classification model that fully utilizes the deep extraction and fusion of physiological features, motion features, and cross-modal features to achieve high-precision classification of the first physiological data and the first motion data obtained by a wearable device. The multi-modal model can keenly capture subtle changes in the user's state and, by learning the patterns of historical normal and abnormal data, enable the classification layer to make accurate probability judgments on real-time data. When the abnormal probability exceeds a preset threshold, the model can promptly identify the abnormal wearing state, avoid triggering information reminders when different users wear the device, and thus perform effective user verification. The combination of the independent feature processing layer and the self-attention mechanism can deeply explore the correlation between features and enhance the model's expressive ability for complex input data. The feature fusion layer further unifies the multi-modal features and improves the generalization performance of the model. Finally, the classification layer compresses and processes the fused feature vectors to accurately calculate the abnormal probability of the device, greatly improving the accuracy and reliability of identifying the abnormal wearing state, helping to accurately identify abnormal wearable devices, and optimizing the overall user experience.

[0055] 2. By considering the specified modes of the device (such as silent mode, do not disturb mode), historical device usage frequency, current battery level, device function relevance, user preferences, signal strength, and the current state of the device, the present invention ensures the comprehensiveness and accuracy of the priority score. It can dynamically select the most suitable device for notification reminders according to the user's actual usage habits and the current state of the device, avoiding interference from multiple devices sending notifications simultaneously. At the same time, it ensures that when the battery of an important device is low or the signal is poor, the notification device can be automatically adjusted, which helps to improve the usage efficiency of the device. This method effectively avoids the situation of irrelevant devices disturbing the user, optimizes the notification delivery method, makes the notifications more timely, accurate, and in line with the user's preferences, thereby enhancing the user experience.

[0056] 3. The present invention dynamically adjusts the message notification method based on historical environmental data and user response time, achieving intelligent and personalized design of the notification method. By recording the environmental data (such as noise level and light intensity) and user response time after notification reminders, it can select the most effective message notification method under similar environmental conditions, thereby optimizing the user's response efficiency. It can automatically select a more suitable notification method for the current environment according to the actual feedback of the user in different environments, avoiding notification neglect or delay caused by environmental discomfort, and improving the timeliness of message notification and the user experience. Especially when the notification method is not specified in advance, it can intelligently match the notification method with the fastest user response in the historically similar environment, ensuring that the message can be conveyed to the user in the most appropriate way, greatly enhancing the adaptability of the method and the user's satisfaction. BRIEF DESCRIPTION OF THE DRAWINGS

[0057] Figure 1 It is a flowchart of a wearable state recognition method provided by an embodiment of the present invention;

[0058] Figure 2 It is a structural diagram of a multi-modal classification model provided by an embodiment of the present invention;

[0059] Figure 3 It is a structural diagram of a wearable state recognition device provided by an embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0060] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0061] Embodiment 1

[0062] Company A launched a mobile application (APP) for wearable devices. The APP aims to provide users with intelligent wearable device management services to help users optimize the usage experience of their smart watches, smart bracelets, and smart rings, etc. Designated user U1 installed the mobile application on their mobile phone, and this mobile phone is the designated mobile phone in this embodiment. The designated mobile phone is currently simultaneously matched and connected with a smart headset, a smart bracelet, Smart Watch 1, Smart Watch 2, and a smart ring; user U1 is currently wearing a smart bracelet, Smart Watch 1, and a smart ring; the smart bracelet, smart watch, and smart ring all have physiological feature recording functions, motion recording functions, environmental data recording functions, and mobile phone synchronization notification functions.

[0063] A wearable state recognition method includes:

[0064] Refer to Figure 1 S10 in

[0065] to obtain the real-time notification status of the user-specified mobile phone to determine whether a new message is received;

[0066] Furthermore, obtaining the real-time notification status of the specified mobile phone includes:

[0067] Continuously monitor the notification center of the specified mobile phone; when it is detected that the specified mobile phone receives a new message, extract the basic information of the new message, including the message type, content summary, source application, reception timestamp, and read status; the read status includes the read status and the unread status; screen the basic information according to the preset data synchronization rules, determine the new messages whose basic information conforms to the data synchronization rules as having new messages received, and update the real-time notification status to yes. According to the pre-set rules, screen the basic information of the notification messages received from the specified mobile phone to decide which notifications should be pushed to the wearable device and which notifications can be ignored (such as advertisement push notifications). The data synchronization rules in this embodiment are user-defined to ensure that the user only receives notifications related to their needs or preferences.

[0068] Refer to Figure 1 S10 in

[0069] If the real-time notification status is yes, execute steps one to five.

[0070] Refer to Figure 1In S11, Step 1: Obtain the set of wearable devices currently paired with the specified mobile phone. The set of wearable devices includes N wearable devices with physiological feature recording functions, exercise recording functions, environmental data recording functions, and mobile phone synchronization notification functions; N ≥ 2. In this embodiment, the set of wearable devices currently paired with the specified mobile phone is obtained. The set of wearable devices includes 5 wearable devices with physiological feature recording functions, exercise recording functions, environmental data recording functions, and mobile phone synchronization notification functions.

[0071] In this embodiment, all wearable devices currently paired with the user-specified mobile phone are obtained through a Bluetooth interface or other communication protocols. This set of wearable devices includes 5 devices, as shown in Table 1. Smart Watch 1 and Smart Watch 2 are two smart watches of different brands.

[0072] Table 1. List of all wearable devices currently paired with the user-specified mobile phone

[0073] Device Name Device Number Smart Bracelet Device 1 Smart Watch 1 Device 2 Smart Ring Device 3 Smart Watch 2 Device 4 Smart Headphones Device 5

[0074] These devices all have physiological feature recording, exercise recording, environmental data recording, and mobile phone synchronization notification functions.

[0075] Refer to Figure 1 In S12, Step 2: Obtain the sensor acquisition data through each of the wearable devices; if the sensor acquisition data meets the specified conditions, determine that the corresponding wearable device is in a worn state; obtain all the wearable devices in the worn state to obtain the worn devices; all the worn devices form a set of worn devices; for each of the worn devices, obtain the corresponding real-time physiological data and real-time exercise data to obtain the first physiological data and the first exercise data; specifically: for each device detected to be in the worn state, obtain the corresponding real-time physiological data and real-time exercise data of the device. Each of the worn devices corresponds to a set of real-time physiological data and a set of real-time exercise data.

[0076] Specifically: if the sensor acquisition data meets the specified conditions, mark the wearable device corresponding to the sensor acquisition data that meets the specified conditions as a worn device;

[0077] Further, obtain the sensor acquisition data of the user through each of the wearable devices; if the sensor acquisition data meets the specified conditions, determining that the corresponding wearable device is in the worn state is specifically:

[0078] The sensor acquisition data includes a detected resistance value, a detected temperature value, and a detected pressure value;

[0079] For each of the wearable devices, the resistance value at the data acquisition location of the wearable device is detected by a skin resistance sensor to obtain the detected resistance value; if the detected resistance value is within a preset resistance range (as shown in Table 2), it passes the skin resistance test; the temperature at the contact point between each wearable device and the skin is measured by a temperature sensor to obtain the detected temperature value; if the detected temperature value is within a preset temperature range (as shown in Table 2), it passes the temperature test; the pressure value at the data acquisition location of each wearable device is detected by a pressure sensor to obtain the detected pressure value; if the detected pressure value is within a preset pressure range (as shown in Table 2), it passes the pressure test;

[0080] Table 2. List of Normal Ranges for Each Sensor Device

[0081]

[0082] The wearable devices that pass the skin resistance test, the temperature test, and the pressure test simultaneously are obtained to get the correctly worn devices; it is determined that the correctly worn devices are in the wearing state.

[0083] Table 3. Wearable Device Sensor Data Acquisition Wear State Identification Table

[0084]

[0085] In this embodiment, the sensor-collected data is as shown in Table 3. Devices 1, 2, 3, and 4 have all passed all sensor tests, including resistance, temperature, and pressure tests, indicating that these devices are correctly worn and are determined to be worn devices. The resistance values, temperature values, and pressure values of each device are within the preset normal ranges. The sensor detection data of Device 5 shows that its resistance value is 10 kΩ, temperature is 25 °C, and pressure is 0 kPa. These data indicate that the device is not in contact with the skin and fails the resistance, temperature, and pressure tests, and is determined to be an unworn device.

[0086] For each of the worn devices (Devices 1, 2, 3, and 4), the corresponding real-time physiological data and real-time motion data are obtained to get the first physiological data and the first motion data, as shown in Table 4;

[0087] Table 4. Real-Time Physiological Data and Real-Time Motion Data of Worn Devices

[0088]

[0089] By performing multi-sensor detection on the skin resistance, temperature, and pressure values of the wearable device, it is ensured that the device can accurately identify whether it is correctly worn. Through the collaborative work of the skin resistance sensor, temperature sensor, and pressure sensor, it can effectively prevent the device from misjudging the wearing state when not worn or worn improperly, and improve the accuracy of device identification. In addition, by using preset resistance, temperature, and pressure ranges for verification, the reliability of sensor data can be ensured, reducing false alarms and missed alarms, thereby improving the accuracy of device wearing state judgment, optimizing the user experience, and ensuring that the wearable device triggers corresponding functions and notification reminders only when it is correctly worn.

[0090] Refer to Figure 1 S13 in, Step 3: Identify the normally worn devices and abnormally worn devices based on the first physiological data and the first motion data respectively; Remove the abnormally worn devices from the set of worn devices to obtain a set of notification receiving devices;

[0091] Further, identifying the normally worn devices and the abnormally worn devices based on the first physiological data and the first motion data includes:

[0092] Obtain the historical normal physiological data and historical normal motion data corresponding to each wearable device in the set of wearable devices; The historical normal physiological data is various physiological parameters recorded by the user when normally wearing the wearable device; The historical normal motion data is various motion parameters recorded by the user when normally wearing the wearable device;

[0093] Utilize historical abnormal records to obtain the physiological data and motion data generated by the user when abnormally wearing the wearable device, to obtain historical abnormal physiological data and historical abnormal motion data;

[0094] Label the historical normal physiological data and the historical normal motion data as positive samples to obtain positive sample labels; Label the historical abnormal physiological data and the historical abnormal motion data as negative samples to obtain negative sample labels; Integrate the positive and negative sample data into a labeled historical physiological data set and a labeled historical motion data set respectively;

[0095] Perform timestamp synchronization and standardization processing on the labeled historical physiological data and the labeled historical motion data, and remove noise, outliers, and fill in missing values by cleaning the data to obtain preprocessed historical physiological data and preprocessed historical motion data;

[0096] Feature extraction is performed on the preprocessed historical physiological data and the preprocessed historical motion data to obtain physiological features, motion features, and cross-modal features; the physiological features include skin conductivity; the motion features include step frequency, step amplitude, step speed, acceleration change rate, motion direction change, and micro-motion features; the cross-modal features include heart rate step frequency ratio and acceleration skin conductivity ratio; the cross-modal features are extracted by combining data of different modalities (physiological and motion data), involving cross-modal feature engineering:

[0097] Among them, the heart rate step frequency ratio is calculated by the ratio of heart rate and step frequency data; the acceleration skin conductivity ratio is calculated by the ratio of acceleration data and skin conductivity data.

[0098] Using the physiological features, the motion features, and the cross-modal features as input data, and using the positive sample label and the negative sample label as training labels, a multi-modal classification model is trained;

[0099] Inputting the first physiological data and the first motion data into the trained multi-modal classification model to obtain an anomaly probability;

[0100] If the anomaly probability exceeds a preset threshold, the wearable device corresponding to the first physiological data and the first motion data is identified as the abnormal wearable device;

[0101] If the anomaly probability does not exceed the preset threshold, the wearable device corresponding to the first physiological data and the first motion data is identified as the normal wearable device.

[0102] Next, the detection of the abnormal wearable device in this embodiment will be described:

[0103] In this embodiment, it is specified that user U1 and user U2 are in the same room. A smart watch (Smart Watch 2) of user U1 has been paired with the designated mobile phone of user U1, but this smart watch is temporarily given to user U2 for use. During the period when user U2 is using it, a new message is received by the designated mobile phone. According to the first physiological data and the first motion data collected by Smart Watch 2, inputting them into the multi-modal classification model, the obtained anomaly probability is 0.92, which exceeds the preset threshold of 0.50. The Smart Watch 2 corresponding to device 4 is identified as the abnormal wearable device; the abnormal wearable device is removed from the set of worn devices to obtain a notification receiving device set; at this time, Smart Watch 2 should be removed from the set of worn devices, and no mobile phone message synchronization reminder is given to user U2.

[0104] To ensure accurate identification and efficient operation of wearable devices during use, and to address issues such as misjudgment and accidental triggering. Through multimodal analysis that combines physiological and motion data, the model can determine the wearing status of the device from multiple dimensions, effectively identify normal and abnormal wear, and prevent the device from accidentally triggering functions when not worn correctly or worn by others. With the accumulation of historical data and positive and negative sample annotation, the model can learn the normal wearing characteristics of users in a personalized manner, improving the accuracy of identification. Through feature extraction of physiological features, motion features, and cross-modal features (such as heart rate step frequency ratio and acceleration skin conductance ratio), different-dimensional data can be fully utilized for comprehensive judgment. This method effectively improves the accuracy of anomaly detection and reduces the possibility of misjudgment and missed judgment. At the same time, through intelligent and automated design, the model can adapt to users' behavioral changes in real time and dynamically without user intervention, ensuring that the device responds in a timely manner in complex environments and optimizing the user experience. In addition, the model can quickly identify abnormal wearing situations, avoid misrecording physiological data or misusing device functions, and thus protect user privacy and device security.

[0105] Furthermore, as Figure 2 shown, the multimodal classification model includes:

[0106] An input layer for receiving input data including the physiological features, the motion features, and the cross-modal features; the input layer contains three sub-input modules for receiving and processing the physiological features, the motion features, and the cross-modal features respectively;

[0107] A feature processing layer including an independent feature processing layer and a self-attention mechanism layer. The independent feature processing layer consists of three fully connected layers, each layer containing 128, 64, and 32 neurons respectively, and using the ReLU activation function. Dropout is applied between each layer to deeply extract the physiological features, the motion features, and the cross-modal features to obtain physiological depth features, motion depth features, and cross-modal depth features; the self-attention mechanism layer is used to weight the physiological depth features, the motion depth features, and the cross-modal depth features to obtain weighted features;

[0108] A feature fusion layer for fusing the weighted features to generate a unified feature vector;

[0109] A classification layer including three fully connected layers for gradually compressing the feature vector generated by the feature fusion layer. Each fully connected layer contains 32, 16, and 1 neuron respectively, and performs non-linear transformation through the ReLU activation function; the last fully connected layer of the classification layer uses the Sigmoid activation function to generate the anomaly probability.

[0110] This multi-modal classification model processes physiological features, motion features, and cross-modal features by inputting them into sub-modules respectively. By combining an independent feature processing layer and a self-attention mechanism layer, the model can extract deep features from multi-dimensional data, improving the recognition accuracy of normal and abnormal wearable devices. The self-attention mechanism further optimizes the correlation between features, ensuring that the model can more accurately focus on key features. The design of the feature fusion layer enables effective fusion of data from different modalities, generating a unified feature vector and enhancing the overall performance of the model. In the classification layer, by gradually compressing the feature vector and using the Sigmoid activation function to output the abnormal probability, the model can accurately judge the wearing state of the device, reduce false alarms and missed detections, and improve the intelligent judgment ability of the device and the user experience.

[0111] Refer to Figure 1 in S14, Step 4: Set a priority for the normal wearable device;

[0112] Refer to Figure 1 in S15, Step 5: Divide the notification receiving device set into a notification device set and a silent device set according to the priority; the notification device set includes only one of the normal wearable devices for notification reminder; the silent device set includes at least one of the normal wearable devices for maintaining a silent state;

[0113] Further, dividing the notification receiving device set into the notification device set and the silent device set according to the priority is specifically:

[0114] If the normal wearable device in the receiving device set is in a specified mode, set the priority score of the normal wearable device in the receiving device set in the specified mode to 0; the specified mode includes a silent mode and a do not disturb mode;

[0115] If the normal wearable device in the receiving device set is not in the specified mode, obtain the historical device usage frequency, current battery level, device function correlation index, user preference index, signal strength score, and device current state of the normal wearable device; the device current state includes an active state and an inactive state; calculate the priority score according to the historical device usage frequency, the current battery level, the device function correlation index, the user preference index, the signal strength score, and the device current state:

[0116] In this embodiment, priorities are set for the normal wearable devices in the notification receiving device set; if a device in the receiving device set is in the silent mode, its priority score is set to 0; if the device is not in the silent mode, the historical device usage frequency, the current battery level, the device function correlation index, the user preference index, the signal strength score, and the current device status of different normal wearable devices in the notification receiving device set are obtained, and the following formula is used to calculate the priority score:

[0117]

[0118] Where PS represents the priority score, F represents the historical device usage frequency, B represents the current battery level, RI represents the device function correlation index, UI represents the user preference index, PI represents the signal strength score, DS represents the current device status, and ω 1 represents the weight of the historical device usage frequency, ω 2 represents the weight of the current battery level, ω 3 represents the weight of the device function correlation index, ω 4 represents the weight of the user preference index, ω 5 represents the weight of the signal strength score, ω 6 represents the weight of the current device status; the current device status includes the device screen-on status (the current device status is 1) and the device screen-off status (the current device status is 0); among them, the historical device usage frequency, the current battery level, the device function correlation index, the user preference index, the signal strength score, and the current device status are all values between 0 and 1.

[0119] All the normal wearable devices in the receiving device set are sorted according to the priority score. As shown in Table 5, the normal wearable device with the highest priority score is selected as the highest-priority device; the highest-priority device is assigned to the notification device set, and the normal wearable devices in the receiving device set except the highest-priority device are assigned to the silent device set. In this embodiment, Device 1 - the smart bracelet is the highest-priority device for message notification.

[0120] Table 5. Priority list of normal wearable devices in the notification receiving device set

[0121] Device Number Device Name Priority Score Priority Ranking Device 1 Smart Bracelet 0.98 1 Device 2 Smart Watch 1 0.88 2 Device 3 Smart Ring 0.60 3

[0122] In this embodiment, when a device is in an inactive state or has insufficient power, the priority score will be correspondingly reduced, and this method will automatically select a more suitable device for notification reminder to ensure the continuity and reliability of the user experience.

[0123] By comprehensively considering the usage frequency of the device, battery power, functional relevance, user preferences, signal strength, and the current state of the device, the wearable device can be intelligently prioritized to ensure that notifications can be promptly pushed to the most suitable device. For devices in the silent mode or do not disturb mode, their priorities are automatically set to 0 to ensure compliance with the user's current usage intention, avoid unnecessary interference, and improve the user experience. By dynamically evaluating the active state and other parameters of the device, the device with the highest priority can be accurately selected for notification reminder, and other devices are placed in the silent state to ensure that the notification is only displayed on the device with the highest priority. This mechanism not only reduces the interference of multiple device simultaneous reminders, but also optimizes the notification delivery efficiency and improves the intelligent management level of the device.

[0124] Further, dividing the set of receiving devices into the set of notification devices and the set of silent devices according to the priority further includes: if the number of devices with the highest priority is greater than one, screening the devices with the highest priority based on a preset rule to determine the final notification device, and allocating the final notification device to the set of notification devices.

[0125] If the number of devices with the highest priority is greater than one, that is, there are normal wearable devices with the same priority score. The preset rules include: preferentially selecting devices in the active state according to whether the device is in the active state (such as the screen is lit, the device is being used); selecting a device with a higher battery power for notification push; screening according to the user's recent usage frequency of the device, and preferentially selecting the device that the user has frequently used recently; selecting a device with a higher signal strength for notification push according to the signal strength between the device and the main device (such as a mobile phone).

[0126] By further screening the devices with the highest priority, it is ensured that when multiple devices have the same highest priority, the most suitable device can be selected as the final notification device according to the preset rules. It is possible to avoid multiple devices receiving notifications simultaneously, reduce redundant notification reminders, and avoid the problem of the user being repeatedly disturbed in a multi-device environment. By setting preset rules, the optimal device can be more accurately determined to ensure that the notification is pushed to the device on which the user is most likely to respond. This not only optimizes the effectiveness of notification delivery, but also improves the personalization and intelligence of the user experience, further reducing notification interference in a multi-device environment and ensuring that the user can obtain key information in a timely manner without being disturbed.

[0127] Further, the method further includes: after the wearable device in the notification device set performs the notification reminder, recording historical environmental data and user response time; the historical environmental data is environmental data collected at a first time corresponding to the notification reminder, including noise level and light intensity; the response time is the time interval from the first time to the detection of the user feedback operation; the user feedback operation includes performing a specified interaction operation through a device in the notification device set or the specified mobile phone;

[0128] When the message notification method is not pre-specified, adjusting the message notification method according to the historical environmental data and the user response time, including:

[0129] When the specified mobile phone receives a new message, obtaining environmental data of the current environment through the wearable device in the notification device set to obtain real-time environmental data, calculating the similarity score between the real-time environmental data and each piece of the historical environmental data, identifying the historical environmental data with the highest similarity score to obtain the first historical environmental data; obtaining the user response time corresponding to the first historical environmental data, and selecting the historical message notification method with the shortest user response time as the message notification method.

[0130] By dynamically adjusting the message notification method by using historical environmental data and user response time, it is ensured that the notification method is more adapted to the current environment and the actual reaction speed of the user. Specifically, by comparing the real-time collected environmental data (such as noise level and light intensity) with the historical environmental data, the historical scenario most similar to the current environment is intelligently identified, and the notification method with the fastest user response in this scenario is selected. It can effectively improve the timeliness and adaptability of notifications, ensuring that users can receive notifications quickly and effectively in different environments.

[0131] In addition, this method also enhances the personalization of the user experience and avoids unnecessary interference. For example, in an environment with high noise, vibration or visual notification can be selected to ensure that the notification is not ignored; while in a dim environment, silent or soft brightness reminder may be preferred. By comprehensively considering the user's past reaction patterns in similar environments, the notification method can be automatically optimized according to the user's actual feedback, reducing the delay or annoyance caused by inappropriate notification methods, thereby enhancing the intelligent level and user satisfaction of the method.

[0132] By intelligently identifying and screening the wearing status of wearable devices, it effectively avoids the interference caused by simultaneous notifications from multiple devices, improves the accuracy of notification reminders and the user experience. By real-time monitoring of physiological data and motion data, it can accurately identify devices in normal and abnormal wearing states, ensuring that notifications are only sent to devices in normal wearing states, thereby enhancing the reliability and practicality of the method. At the same time, by setting priorities for devices in normal wearing states, it further optimizes the distribution method of notifications, reduces unnecessary notification interference, ensures the timely delivery of important notifications, and helps to improve the user experience.

[0133] Embodiment 2

[0134] Company B has launched an intelligent wearing status recognition and notification management system based on wearable devices, specifically a wearing status recognition device. Users use this device to manage multiple wearable devices, including smart bracelets, smart watches, smart rings and other devices. This device can identify the correct wearing status of the devices, collect physiological and motion data, and send notifications to the devices according to priorities. A certain user has applied a wearing status recognition and device.

[0135] A wearing status recognition device, as Figure 3 shown, includes:

[0136] A notification status acquisition module, used to acquire the real-time notification status of a user-specified mobile phone to determine whether a new message is received;

[0137] A wearable device set acquisition module, used to acquire the set of wearable devices currently paired with the specified mobile phone, where the set of wearable devices includes N wearable devices with physiological feature recording functions, motion recording functions, environmental data recording functions, and mobile phone synchronization notification functions; N≥2;

[0138] A data acquisition module, used to acquire sensor acquisition data through each of the wearable devices; if the sensor acquisition data meets the specified conditions, it is determined that the corresponding wearable device is in a wearing state; all the wearable devices in the wearing state are acquired to obtain the worn devices; all the worn devices form a set of worn devices; for each of the worn devices, the corresponding real-time physiological data and real-time motion data are acquired to obtain first physiological data and first motion data;

[0139] An abnormal device identification module, used to identify normal wearing devices and abnormal wearing devices respectively according to the first physiological data and the first motion data; the abnormal wearing devices are removed from the set of worn devices to obtain a set of notification receiving devices;

[0140] A priority setting module, used to set priorities for the normal wearing devices;

[0141] The device allocation module is used to divide the set of notification receiving devices into a notification device set and a silent device set according to the priority; the notification device set only includes one of the normal wearable devices for notification reminder; the silent device set includes at least one of the normal wearable devices for maintaining a silent state;

[0142] The control module is used to control the device to execute the operations of the wearable device set acquisition module, the data acquisition module, the abnormal device identification module, the priority setting module, and the device allocation module when the notification status acquisition module detects that the real-time notification status is yes.

[0143] Specifically, as Figure 1 shown, if the real-time notification status is yes, the following steps are executed, including:

[0144] Step 1: Obtain the set of wearable devices currently paired with the specified mobile phone. The set of wearable devices includes N wearable devices with physiological feature recording functions, motion recording functions, environmental data recording functions, and mobile phone synchronous notification functions;

[0145] Step 2: Obtain sensor acquisition data through each wearable device; if the sensor acquisition data meets the specified conditions, determine that the corresponding wearable device is in a worn state; obtain all the wearable devices in the worn state to obtain the set of worn devices; for each of the worn devices, obtain the corresponding real-time physiological data and real-time motion data to obtain the first physiological data and the first motion data;

[0146] Step 3: Identify the normal wearable devices and abnormal wearable devices respectively according to the first physiological data and the first motion data; remove the abnormal wearable devices from the set of worn devices to obtain the set of notification receiving devices;

[0147] Step 4: Set priorities for the normal wearable devices;

[0148] Step 5: Divide the set of notification receiving devices into a notification device set and a silent device set according to the priority; the notification device set only includes one of the normal wearable devices for notification reminder; the silent device set includes at least one of the normal wearable devices for maintaining a silent state.

[0149] Furthermore, obtaining the real-time notification status of the specified mobile phone includes:

[0150] Continuously monitor the notification center of the specified mobile phone; when it is detected that the specified mobile phone receives a new message, extract the basic information of the new message, including the message type, content summary, source application, reception timestamp, and read status; screen the basic information according to the preset data synchronization rules, determine that the new message whose basic information conforms to the data synchronization rules has a new message received, and update the real-time notification status to yes.

[0151] Further, obtain the sensor acquisition data of the user through each of the wearable devices; if the sensor acquisition data meets the specified conditions, determining that the corresponding wearable device is in the worn state specifically includes:

[0152] The sensor acquisition data includes a detected resistance value, a detected temperature value, and a detected pressure value;

[0153] For each of the wearable devices, detect the resistance value at the data acquisition location of the wearable device through a skin resistance sensor to obtain the detected resistance value; if the detected resistance value is within the preset resistance range, pass the skin resistance test; measure the temperature at the skin contact point of each of the wearable devices through a temperature sensor to obtain the detected temperature value; if the detected temperature value is within the preset temperature range, pass the temperature test; detect the pressure value at the data acquisition location of each of the wearable devices through a pressure sensor to obtain the detected pressure value; if the detected pressure value is within the preset pressure range, pass the pressure test;

[0154] Obtain the wearable devices that pass the skin resistance test, the temperature test, and the pressure test simultaneously to obtain the correctly worn devices; determine that the correctly worn devices are in the worn state.

[0155] Further, identifying the normal wearable devices and the abnormal wearable devices according to the first physiological data and the first motion data includes:

[0156] Obtain the historical normal physiological data and historical normal motion data corresponding to each of the wearable devices in the wearable device set; the historical normal physiological data are various physiological parameters recorded by the user when the wearable device is worn normally; the historical normal motion data are various motion parameters recorded by the user when the wearable device is worn normally;

[0157] Utilize the historical abnormal records to obtain the physiological data and motion data generated by the user when the wearable device is worn abnormally, to obtain historical abnormal physiological data and historical abnormal motion data;

[0158] Label the historical normal physiological data and the historical normal motion data as positive samples to obtain positive sample labels; label the historical abnormal physiological data and the historical abnormal motion data as negative samples to obtain negative sample labels; integrate the positive and negative sample data into a labeled historical physiological data set and a labeled historical motion data set respectively;

[0159] Perform timestamp synchronization and normalization processing on the labeled historical physiological data and the labeled historical motion data, and remove noise, outliers, and fill in missing values by cleaning the data to obtain preprocessed historical physiological data and preprocessed historical motion data;

[0160] Extract features from the preprocessed historical physiological data and the preprocessed historical motion data to obtain physiological features, motion features, and cross-modal features; the physiological features include skin conductivity; the motion features include step frequency, step amplitude, step speed, acceleration change rate, motion direction change, and micro-motion features; the cross-modal features include heart rate-step frequency ratio and acceleration-skin conductivity ratio;

[0161] Use the physiological features, the motion features, and the cross-modal features as input data, and use the positive sample labels and the negative sample labels as training labels to train a multi-modal classification model;

[0162] Input the first physiological data and the first motion data into the trained multi-modal classification model to obtain an abnormal probability;

[0163] If the abnormal probability exceeds a preset threshold, identify the wearable device corresponding to the first physiological data and the first motion data as the abnormal wearable device;

[0164] If the abnormal probability does not exceed the preset threshold, identify the wearable device corresponding to the first physiological data and the first motion data as the normal wearable device.

[0165] Further, the multi-modal classification model includes:

[0166] An input layer for receiving input data including the physiological features, the motion features, and the cross-modal features; the input layer contains three sub-input modules for receiving and processing the physiological features, the motion features, and the cross-modal features respectively;

[0167] The feature processing layer includes an independent feature processing layer and a self-attention mechanism layer. The independent feature processing layer consists of three fully-connected layers, each containing 128, 64, and 32 neurons respectively, and uses the ReLU activation function. Dropout is applied between each layer to deeply extract the physiological features, the motion features, and the cross-modal features, obtaining physiological depth features, motion depth features, and cross-modal depth features. The self-attention mechanism layer is used to weight the physiological depth features, the motion depth features, and the cross-modal depth features to obtain weighted features.

[0168] The feature fusion layer is used to fuse the weighted features to generate a unified feature vector.

[0169] The classification layer includes three fully-connected layers, which are used to gradually compress the feature vector generated by the feature fusion layer. Each fully-connected layer contains 32, 16, and 1 neuron respectively, and performs non-linear transformation through the ReLU activation function. The last fully-connected layer of the classification layer uses the Sigmoid activation function to generate the anomaly probability.

[0170] Further, dividing the notification receiving device set into the notification device set and the silent device set according to the priority specifically is as follows:

[0171] If the normal wearable devices in the receiving device set are in the specified mode, set the priority score of the normal wearable devices in the receiving device set in the specified mode to 0. The specified mode includes the silent mode and the do-not-disturb mode.

[0172] If the normal wearable devices in the receiving device set are not in the specified mode, obtain the historical device usage frequency, the current battery power, the device function correlation index, the user preference index, the signal strength score, and the device current state of the normal wearable devices. The device current state includes the active state and the inactive state. Calculate the priority score according to the historical device usage frequency, the current battery power, the device function correlation index, the user preference index, the signal strength score, and the device current state:

[0173] Sort all the normal wearable devices in the receiving device set according to the priority score, select the normal wearable device with the highest priority score as the highest-priority device, assign the highest-priority device to the notification device set, and assign the normal wearable devices in the receiving device set except the highest-priority device to the silent device set.

[0174] Further, dividing the set of receiving devices into the set of notification devices and the set of silent devices according to the priority further includes: if the number of the devices with the highest priority is greater than one, screening the devices with the highest priority based on a preset rule to determine the final notification devices, and allocating the final notification devices to the set of notification devices.

[0175] Further, the method further includes: after the wearable device in the set of notification devices performs the notification reminder, recording historical environment data and the user response time; the historical environment data is the environment data collected at a first time corresponding to the notification reminder, including the noise level and the light intensity; the response time is the time interval from the first time to the time when a user feedback operation is detected; the user feedback operation includes performing a specified interaction operation through a device in the set of notification devices or the specified mobile phone.

[0176] When the message notification method is not specified in advance, adjusting the message notification method according to the historical environment data and the user response time, including:

[0177] When the specified mobile phone receives a new message, obtaining the environment data of the current environment through the wearable device in the set of notification devices to obtain real-time environment data, calculating the similarity score between the real-time environment data and each piece of the historical environment data, identifying the historical environment data with the highest similarity score to obtain the first historical environment data; obtaining the user response time corresponding to the first historical environment data, and selecting the historical message notification method with the shortest user response time as the message notification method.

[0178] Embodiment III

[0179] A computer-readable storage medium, on which a computer program is stored, and when the computer program is executed by a processor, the steps of the method described in Embodiment 1 of the present disclosure are implemented.

[0180] Although the embodiments of the present invention have been shown and described, for those of ordinary skill in the art, it can be understood that various changes, modifications, substitutions and variations can be made to these embodiments without departing from the principle and spirit of the present invention, and the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A wearing status recognition method, characterized in that: include: Get the real-time notification status of the user's specified mobile phone to determine whether new messages have been received; If the real-time notification status is yes, the following steps are performed, specifically including: Step 1: Obtain a set of wearable devices currently paired with the specified mobile phone, wherein the set of wearable devices includes N wearable devices having a physiological characteristic recording function, a motion recording function, an environmental data recording function, and a mobile phone synchronization notification function; Step 2: acquiring sensor data through each of the wearable devices; if the sensor data meets the specified conditions, determining that the corresponding wearable device is in a wearing state; acquiring all the wearable devices in the wearing state to obtain a worn device; all the worn devices constitute a worn device set; for each of the worn devices, acquiring corresponding real-time physiological data and real-time motion data to obtain first physiological data and first motion data; Step 3: identifying normal wearable devices and abnormal wearable devices according to the first physiological data and the first motion data respectively; removing the abnormal wearable devices from the set of worn devices to obtain a set of notification receiving devices; Step 4: Setting a priority for the normal wearable device; Step 5: Divide the notification receiving device set into a notification device set and a silent device set according to the priority; the notification device set includes only one normal wearable device for notification reminders; the silent device set includes at least one normal wearable device for maintaining a silent state.

2. A wearing status recognition method according to claim 1, characterized in that: Acquiring the real-time notification status of the designated mobile phone includes: Continuously monitor the notification center of the designated mobile phone; when it is detected that the designated mobile phone has received a new message, extract the basic information of the new message, including the message type, content summary, source application, receiving timestamp and reading status; filter the basic information according to preset data synchronization rules, judge the new message whose basic information meets the data synchronization rules as a new message received, and update the real-time notification status to yes.

3. A wearing status recognition method according to claim 1, characterized in that: Acquiring the sensor collected data of the user through each wearable device; if the sensor collected data meets the specified condition, determining that the corresponding wearable device is in the wearing state is specifically: The sensor collects data including detection resistance value, detection temperature value and detection pressure value; For each of the wearable devices, a resistance value at a data collection point of the wearable device is detected by a skin resistance sensor to obtain the detected resistance value; if the detected resistance value is within a preset resistance range, the skin resistance test is passed; Measuring the temperature of each contact point between the wearable device and the skin by a temperature sensor to obtain the detected temperature value; If the detected temperature value is within the preset temperature range, the temperature test is passed; Detecting the pressure value at each wearable device data collection point through a pressure sensor to obtain the detected pressure value; if the detected pressure value is within a preset pressure range, the pressure test is passed; Acquire the wearable device that passes the skin resistance test, the temperature test, and the pressure test at the same time, and obtain a correct wearable device; Determine that the correct wearable device is in the wearing state.

4. A wearing status recognition method according to claim 1, characterized in that: Identifying the normal wearable device and the abnormal wearable device according to the first physiological data and the first motion data includes: Acquire historical normal physiological data and historical normal motion data corresponding to each wearable device in the wearable device set; the historical normal physiological data are various physiological parameters recorded when the user normally wears the wearable device; the historical normal motion data are various motion parameters recorded when the user normally wears the wearable device; Acquire the physiological data and motion data generated by the user when abnormally wearing the wearable device by using the historical abnormal records, and obtain the historical abnormal physiological data and the historical abnormal motion data; The historical normal physiological data and the historical normal motion data are marked as positive samples to obtain positive sample labels; the historical abnormal physiological data and the historical abnormal motion data are marked as negative samples to obtain negative sample labels; the positive and negative sample data are respectively integrated into a marked historical physiological data set and a marked historical motion data set; Performing timestamp synchronization and standardization processing on the annotated historical physiological data and the annotated historical motion data, and removing noise, outliers, and filling missing values ​​by cleaning the data to obtain preprocessed historical physiological data and preprocessed historical motion data; Feature extraction is performed on the preprocessed historical physiological data and the preprocessed historical motion data to obtain physiological features, motion features and cross-modal features; the physiological features include skin conductivity; the motion features include step frequency, stride length, step speed, acceleration change rate, motion direction change and micro-motion features; the cross-modal features include heart rate step frequency ratio and acceleration skin conductivity ratio; Using the physiological features, the motion features, and the cross-modal features as input data, and using the positive sample labels and the negative sample labels as training labels, to train a multimodal classification model; Inputting the first physiological data and the first motion data into the trained multimodal classification model to obtain an abnormality probability; If the abnormal probability exceeds a preset threshold, identifying the wearable device corresponding to the first physiological data and the first motion data as the abnormal wearable device; If the abnormal probability does not exceed the preset threshold, the wearable device corresponding to the first physiological data and the first motion data is identified as the normal wearable device.

5. A wearing status recognition method according to claim 4, characterized in that: The multimodal classification model includes: An input layer, used for receiving input data including the physiological feature, the motion feature and the cross-modal feature; the input layer comprises three sub-input modules, respectively used for receiving and processing the physiological feature, the motion feature and the cross-modal feature; A feature processing layer, including an independent feature processing layer and a self-attention mechanism layer, wherein the independent feature processing layer is composed of three fully connected layers, each layer contains 128, 64 and 32 neurons respectively, and uses a ReLU activation function, and Dropout is applied between each layer, so as to perform deep extraction on the physiological features, the motion features and the cross-modal features to obtain physiological depth features, motion depth features and cross-modal depth features; the self-attention mechanism layer is used to weight the physiological depth features, the motion depth features and the cross-modal depth features to obtain weighted features; The feature fusion layer is used to fuse the weighted features to generate a unified feature vector; The classification layer includes three fully connected layers, which are used to gradually compress the feature vectors generated by the feature fusion layer, each of which contains 32, 16 and 1 neurons respectively, and performs nonlinear transformation through the ReLU activation function; the last fully connected layer of the classification layer adopts the Sigmoid activation function to generate the abnormality probability.

6. A wearing status recognition method according to claim 1, characterized in that: The notification receiving device set is divided into the notification device set and the silent device set according to the priority: If the normal wearable device in the receiving device set is in a specified mode, setting the priority score of the normal wearable device in the receiving device set in the specified mode to 0; the specified mode includes a silent mode and a do not disturb mode; If the normal wearable device in the receiving device set is not in the specified mode, obtaining the historical device usage frequency, current battery power, device function relevance index, user preference index, signal strength score and current status of the normal wearable device; The current state of the device includes an active state and an inactive state; The priority score is calculated according to the historical device usage frequency, the current battery power, the device function relevance index, the user preference index, the signal strength score, and the current state of the device: All the normal wearable devices in the receiving device set are sorted according to the priority score, and the normal wearable device with the highest priority score is selected as the highest priority device; the highest priority device is assigned to the notification device set, and the normal wearable devices in the receiving device set except the highest priority device are assigned to the silent device set.

7. A wearing status recognition method according to claim 6, characterized in that: Dividing the receiving device set into the notification device set and the silent device set according to the priority also includes: if the number of the highest priority devices is greater than one, screening the highest priority devices based on preset rules, determining the final notification device, and assigning the final notification device to the notification device set.

8. A wearing status recognition method according to claim 1, characterized in that: The method further includes: after the wearable device in the notification device set performs the notification reminder, recording historical environmental data and user response time; the historical environmental data is environmental data collected at the first time corresponding to the notification reminder, including noise level and light intensity; the response time is the time interval from the first time to the detection of the user feedback operation; the user feedback operation includes performing a specified interactive operation through the device in the notification device set or the specified mobile phone; When the message notification mode is not pre-specified, adjusting the message notification mode according to the historical environment data and the user response time includes: When the designated mobile phone receives a new message, the environmental data of the current environment is obtained through the wearable device in the notification device set to obtain real-time environmental data, and the similarity score between the real-time environmental data and each of the historical environmental data is calculated to identify the historical environmental data with the highest similarity score to obtain the first historical environmental data; the user response time corresponding to the first historical environmental data is obtained, and the historical message notification method with the shortest user response time is selected as the message notification method.

9. A wearing status recognition device, characterized in that: include: The notification status acquisition module is used to obtain the real-time notification status of the user's specified mobile phone to determine whether a new message has been received; A wearable device set acquisition module, used to acquire a wearable device set currently paired with the specified mobile phone, wherein the wearable device set includes N wearable devices having a physiological characteristic recording function, a motion recording function, an environmental data recording function, and a mobile phone synchronization notification function; A data acquisition module, used to acquire sensor data through each of the wearable devices; If the data collected by the sensor meets the specified conditions, it is determined that the corresponding wearable device is in a wearing state; Acquire all the wearable devices in the wearing state to obtain the worn devices; All the worn devices form a worn device set; for each of the worn devices, corresponding real-time physiological data and real-time motion data are acquired to obtain first physiological data and first motion data; an abnormal device identification module, used to identify a normal wearable device and an abnormal wearable device according to the first physiological data and the first motion data respectively; The abnormal wearable device is removed from the set of worn devices to obtain a notification receiving device set; A priority setting module, used to set a priority for the normal wearable device; A device allocation module, used for dividing the notification receiving device set into a notification device set and a silent device set according to the priority; the notification device set includes only one normal wearable device for notification reminder; The silent device set includes at least one normal wearable device, which is used to maintain a silent state; A control module is used to control the device to execute the operations of the wearable device set acquisition module, the data acquisition module, the abnormal device identification module, the priority setting module and the device allocation module when the notification status acquisition module detects that the real-time notification status is yes.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 8 are implemented.

Images