Intelligent wearing behavior classification method and system based on AI technology

Through AI technology, analyzing user behavior data and scenes, combining wrist movement postures, smart wearable watches avoid awakening in non-essential scenarios, solving the problem of distraction and rest effects caused by ignoring scene classification in the existing technology, providing more Reasonable experience.

CN119046823AInactive Publication Date: 2024-11-29SHENZHEN TJD INFORMATION TECH CO LTD
View PDF 7 Cites 0 Cited by

Patent Information

Application Number
CN202411533891.7
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-10-31
Publication Date
2024-11-29
Estimated Expiration
Not applicable · inactive patent

AI Technical Summary

Technical Problem

Existing smart wearable behavior classification methods usually focus on market demand and ignore usage scenario classification, which leads to awakening smart watches to distract or affecting rest in non-essential scenarios, causing trouble to users.

Method used

Using an intelligent wearable behavior classification method based on AI technology, users' behavior data are collected through intelligent wearable watches, scene classification model is trained, and the user's current scene is analyzed, and the wrist movement posture is determined whether to wake up. Screen, avoid unnecessary wake-up.

Benefits of technology

It realizes a detailed classification of the user's current usage scenarios, avoids waking up the smartwatch in unnecessary scenarios, reduces the impact on the user's attention distraction and rest, and is easy to use.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119046823A_ABST
    Figure CN119046823A_ABST
Patent Text Reader

Abstract

The invention discloses an intelligent wearing behavior classification method and system based on the AI technology, and belongs to the field of intelligent wearing, and the method comprises the following steps: an intelligent wearing watch collects behavior data of a user, and sends the behavior data to an AI analysis platform; wherein the behavior data comprises real-time physiological sign data, real-time motion data and positioning data; based on the collected previous user behavior data and the corresponding time, training a pre-established classifier to obtain a scene classification model; analyzing the current scene of the user; classifying the current wrist motion posture, and determining whether to wake up a screen of the intelligent wearable watch according to a classification result; and determining whether to update the current scene of the user based on the operation of the user on the intelligent wearable watch after the wrist motion posture classification. According to the method and the device, not only can the current use scene of the user be finely classified, but also whether the screen of the intelligent wearable watch is awakened or not can be analyzed according to the classified use scene.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the field of smart wear, and in particular to a smart wear behavior classification method and system based on AI technology. Background Art

[0002] Smart wearable devices refer to the general term for wearable devices that are developed by applying wearable technology to intelligently design daily wearables, including watches, bracelets, glasses, clothing, etc. Such devices usually have multiple functions, such as health monitoring, sports tracking, daily interaction, etc., and have been widely used in many fields such as medical, industrial, military, education, and entertainment.

[0003] The most common smart wearable device is the smart wearable watch. In order to improve the user experience of wearing smart wearable watches, by classifying smart wearable behaviors, the functions and positioning of different products can be more clearly defined. Classification helps to identify the specific needs of different user groups, thereby guiding product development and marketing. For example, sports enthusiasts may prefer smart wearable watches with rich sports monitoring functions, while the elderly may pay more attention to health monitoring and emergency call functions.

[0004] The existing technology has the following problems: the existing smart wear behavior classification methods usually focus on differentiation based on market demand, but ignore the classification of usage scenarios. It is easy to wake up the smart watch in non-essential scenarios, distracting attention or affecting rest, causing great trouble to users.

[0005] Therefore, those skilled in the art provide a smart wear behavior classification method and system based on AI technology to solve the problems raised in the above background technology. Summary of the invention

[0006] In view of the defects in the prior art, the present invention provides a smart wear behavior classification method based on AI technology, comprising the following steps: The smart wearable watch collects the user's behavior data and sends the behavior data to the AI ​​analysis platform; wherein the behavior data includes real-time physiological sign data, real-time motion data and positioning data; Based on the collected previous user behavior data and the corresponding time, the pre-established classifier is trained to obtain a scene classification model; Based on the collected real-time user behavior data, combined with the current network time and scene classification model, the user's current scene is analyzed; Capture the user's wrist motion posture, classify the current wrist motion posture based on the user's current scene, and decide whether to wake up the screen of the smart wearable watch based on the classification result; Based on the user's operation on the smart wearable watch after the wrist movement posture is classified, decide whether to update the user's current scene.

[0007] As a further solution of the present invention: the specific training process of the scene classification model is: Divide the time corresponding to the previous user behavior data into working days and weekends; The user behavior data belonging to working days are divided into working day clusters, and the user behavior data belonging to rest days are divided into rest day clusters; Divide a day into eight time periods, and organize the user behavior data in the workday cluster and the weekend cluster into corresponding sub-clusters according to the divided time periods; Import the scene information corresponding to the previous user behavior data to add scene tags to the user behavior data in each sub-cluster. The scene tags include the mark of the scene the user is in and whether the do not disturb function is turned on.

[0008] As a further solution of the present invention: the specific analysis process of the scene currently located by the user is: Generate a real-time physiological sign radar chart A based on the collected real-time physiological sign data, wherein the physiological sign data includes heart rate, blood pressure, blood oxygen saturation and skin conductivity; Generate a real-time sports radar chart B based on the collected real-time sports data, wherein the real-time sports data includes step frequency, breathing frequency, calorie consumption and speed; Determine the user's current location P based on the collected positioning data; According to the current network time, determine the sub-cluster of the corresponding time period of the collected real-time user behavior data; Based on the user behavior data in the sub-cluster of the corresponding time period, generate the physiological sign radar chart ai and the movement radar chart bi of the same time period in the past, and the location pi, i=1···n, where n is a positive integer; Match the real-time physiological sign radar chart A, real-time motion radar chart B, and current location P corresponding to the real-time user behavior data with the groups of physiological sign radar charts ai, motion radar charts bi, and location pi in the same time period in the past one by one, and select the ten user behavior data in the same time period in the past that best match the real-time user behavior data; Extract the scene labels corresponding to the ten selected user behavior data in the same time period in the past; Based on the extracted scene tags, the scene tag with the most repetitions is output as the user's current scene.

[0009] As a further solution of the present invention: the smart wearable watch supports a Wi-Fi positioning function. When determining the user's current location, if the user's current location P is the user's residence or workplace, the smart wearable watch further positions the user by identifying the Wi-Fi signal strength of the user's residence or workplace and combining it with the Wi-Fi location database.

[0010] As a further solution of the present invention: the specific process of screening out the ten previous user behavior data of the same time period that best match the real-time user behavior data is as follows: Compare A with the corresponding ai to obtain a first similarity weight value V1; Compare B with the corresponding bi for similarity to obtain a second similarity weight value V2; Match P with the corresponding pi, calculate the straight-line distance between them and mark it as L; Calculate the matching degree S = (V1*V2 / L)*100% between the user behavior data of the same time period in the past and the real-time user behavior data; Filter out the ten user behavior data with the largest S in the same time period.

[0011] As a further solution of the present invention: when the matching degree S is screened, if the ten largest matching degrees S screened out are all less than the preset matching degree, an update event is triggered, the smart wearable watch notifies the user and the user imports the current scene information.

[0012] As a further solution of the present invention: in the process of outputting the scene label with the most repetition times as the user's current scene, if there are multiple scene labels with the most repetition times, the matching degree between the user behavior data of the same time period in the past and the real-time user behavior data corresponding to each scene label is extracted, and then the sum of the matching degrees corresponding to each scene label is calculated, and the scene label with the largest sum of matching degrees is output as the user's current scene.

[0013] As a further solution of the present invention: the specific process of classifying the current wrist movement posture is: Get the user's current scene, that is, the corresponding scene label; The captured wrist motion posture is compared with the four preset wrist postures for similarity. If the similarity of any group of comparisons is greater than a preset value, the next step is entered. If the similarities of the four groups of comparisons are all less than or equal to the preset value, the current wrist motion posture is planned into the first sequence. Extract the scene tag and determine whether the user's current scene is open for Do Not Disturb based on the tag. If Do Not Disturb is turned on, the current wrist movement posture is planned to the second sequence; if Do Not Disturb is not turned on, the current wrist movement posture is planned to the third sequence; If the user's current wrist motion posture is planned to the first sequence or the second sequence, the screen of the smart wearable watch will not be woken up; if the user's current wrist motion posture is planned to the third sequence, the screen of the smart wearable watch will be woken up.

[0014] As a further solution of the present invention: the specific process of updating the state recognition model based on the user's operation on the smart wearable watch after the wrist movement posture classification is: If the current wrist motion posture is planned to the first sequence, and the user repeatedly wakes up the smart wearable watch or manually wakes up the smart wearable watch after the wrist motion posture is classified, the preset wrist posture is updated, that is, one of the four preset wrist postures is randomly replaced with the current user's wrist motion posture, where repeated wake-up means that the number of times the user raises his wrist exceeds the preset value within the preset time period; If the current wrist motion posture is planned to the second sequence, and the user repeatedly wakes up or manually wakes up the smart wearable watch after the wrist motion posture is classified, an update event is triggered, the smart wearable watch notifies the user, and the user imports the current scene information.

[0015] The present application also discloses an AI-based smart wear behavior classification system, which adopts an AI-based smart wear behavior classification method, including a smart wear watch and an AI analysis platform, wherein the AI ​​analysis platform includes a scene classification module, a scene analysis module, a posture classification module, and an update module; The smart wearable watch is used to collect user behavior data and send the behavior data to the AI ​​analysis platform; wherein the behavior data includes real-time physiological sign data, real-time motion data and positioning data; The scene classification module trains a pre-established classifier based on the collected previous user behavior data and the corresponding time to obtain a scene classification model; The scene analysis module analyzes the user's current scene based on the collected real-time user behavior data, combined with the current network time and scene classification model; The posture classification module is used to capture the user's wrist motion posture, classify the current wrist motion posture in combination with the user's current scene, and decide whether to wake up the screen of the smart wearable watch according to the classification result; The update module determines whether to update the user's current scene based on the user's operation on the smart wearable watch after the wrist movement posture is classified.

[0016] The beneficial effects of the present invention are embodied in: This application can not only classify the user's current usage scenarios in detail, but also analyze whether to wake up the screen of the smart wearable watch based on the classified usage scenarios, so as to avoid waking up the smart watch in non-essential scenarios to distract the user's attention or affect his rest, thereby avoiding causing trouble to the user and facilitating use. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] In order to more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the following is a brief introduction to the drawings required for the specific embodiments or the description of the prior art. In all the drawings, similar elements or parts are generally identified by similar reference numerals. In the drawings, the elements or parts are not necessarily drawn according to the actual scale.

[0018] Figure 1 This is a flow chart of a smart wear behavior classification method based on AI technology; Figure 2 A schematic diagram of a scene classification model in a smart wearable behavior classification method based on AI technology; Figure 3 This is a structural block diagram of an intelligent wearable behavior classification system based on AI technology. DETAILED DESCRIPTION

[0019] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0020] As mentioned in the background technology of this application, research has found that existing smart wearable behavior classification methods usually focus on differentiation based on market demand, but ignore the classification of usage scenarios. It is easy to wake up the smart watch in non-essential scenarios, distracting attention or affecting rest, causing great trouble to users, and there are certain defects.

[0021] In order to solve the above-mentioned defects, the present application discloses a smart wear behavior classification method and system based on AI technology, which can not only classify the user's current usage scenarios in detail, but also analyze whether to wake up the screen of the smart wearable watch according to the classified usage scenarios, so as to avoid waking up the smart watch in non-essential scenarios to distract the user's attention or affect his rest, thereby avoiding causing trouble to the user and facilitating use.

[0022] The following will describe in detail how the solution of the present application solves the above technical problems in conjunction with the accompanying drawings.

[0023] See also Figure 1In an embodiment of the present invention, a smart wear behavior classification method based on AI technology includes the following steps: the smart wear watch collects the user's behavior data and sends the behavior data to the AI ​​analysis platform; wherein the behavior data includes real-time physiological sign data, real-time motion data and positioning data; based on the collected previous user behavior data and the corresponding time, the pre-established classifier is trained to obtain a scene classification model; based on the collected real-time user behavior data, combined with the current network time and the scene classification model, the user's current scene is analyzed; the user's wrist movement posture is captured, and the current wrist movement posture is classified in combination with the user's current scene, and whether to wake up the screen of the smart wear watch is determined according to the classification result; based on the user's operation on the smart wear watch after the wrist movement posture is classified, it is determined whether to update the user's current scene. This application can not only classify the user's current usage scenario in detail, but also analyze whether to wake up the screen of the smart wear watch according to the classified usage scenario, so as to avoid waking up the smart watch in non-essential scenarios to distract the user's attention or affect rest, thereby avoiding causing trouble to the user and facilitating use.

[0024] In this embodiment, the specific training process of the scene classification model is as follows: dividing the time corresponding to the previous user behavior data into working days and rest days; dividing the user behavior data belonging to working days into working day clusters, and dividing the user behavior data belonging to rest days into rest day clusters; dividing a day into eight time periods, and planning the user behavior data in the working day cluster and the rest day cluster into corresponding sub-clusters according to the divided time periods; importing the scene information corresponding to the previous user behavior data (the user behavior data corresponding to the user in different scenes is different, for example, the user behavior data in the sports scene and the sleep scene are quite different, and the imported scene information is imported by the user into the AI ​​analysis platform through the smart wearable watch after experiencing the previous scene), so as to add scene labels to the user behavior data in each sub-cluster (such as Figure 2 As shown in the figure, the scene label includes the mark of the scene the user is in and whether the do not disturb function is turned on (for example, the current scene is marked as driving, and the user imports the mark to turn on the do not disturb function; similarly, the current scene is marked as sleeping, and the user also imports the mark to turn on the do not disturb function to prevent the user from accidentally waking up the smart wearable watch screen and affecting driving or sleeping). Each user has his or her own work and rest routine, which is usually not much different during weekdays, but usually has some discrepancies on weekends. This application classifies the user's behavior data according to weekdays and weekends, and then adds corresponding scene labels to each user's behavior data to enrich the database and complete the training.

[0025] In this embodiment, the specific analysis process of the user's current scene is as follows: based on the real-time physiological sign data collected, a real-time physiological sign radar chart A is generated, wherein the physiological sign data include heart rate, blood pressure, blood oxygen saturation and skin conductivity; based on the real-time motion data collected, a real-time motion radar chart B is generated, wherein the real-time motion data include step frequency, breathing frequency, calorie consumption and speed; based on the collected positioning data, the user's current location P is determined; based on the current network time, a sub-cluster of the corresponding time period of the collected real-time user behavior data is determined; based on the user behavior data in the sub-cluster of the corresponding time period, a previous sub-cluster of the same time period is generated The real-time physiological sign radar chart ai, the sports radar chart bi, and the location pi of the segment, i=1···n, where n is a positive integer; the real-time physiological sign radar chart A, the real-time sports radar chart B, and the current location P corresponding to the real-time user behavior data are matched one by one with the physiological sign radar chart ai, the sports radar chart bi, and the location pi of the previous same time period, and the ten previous user behavior data of the same time period that best match the real-time user behavior data are selected; the scene labels corresponding to the ten previous user behavior data of the same time period are extracted; based on the extracted scene labels, the scene label with the most repetitions is output as the user's current scene. Radar chart is a graphical method for displaying multiple quantitative variables. It uses multiple rays (usually representing different variables or dimensions) starting from the same point, and the points marked on these rays (representing the measurement values ​​or scores on the corresponding variables) to show the distribution and changes of data of multiple variables (usually quantitative variables) relative to the center point. Radar charts are mainly used to display the multidimensional distribution characteristics of data and compare the performance of different data points in various dimensions, so as to help users discover patterns and trends in data.

[0026] In this embodiment, the smart wearable watch supports the Wi-Fi positioning function. When determining the current location of the user, if the current location P of the user is the user's residence or workplace, the smart wearable watch further locates by identifying the Wi-Fi signal strength of the user's residence or workplace and combining it with the Wi-Fi location database. The Wi-Fi positioning function is a method of determining the location of a device using the signal information of nearby Wi-Fi hotspots. Its basic principle is that when the device (smart wearable watch) turns on the Wi-Fi function and searches for available networks, it detects nearby Wi-Fi hotspots. Each Wi-Fi hotspot has a unique identifier, called a MAC address. The device collects the MAC address and signal strength (RSSI, i.e., received signal strength index) of nearby hotspots. There is a large database that stores the MAC addresses of Wi-Fi hotspots worldwide and their corresponding geographic locations. When the device needs to perform Wi-Fi positioning, it compares the detected Wi-Fi hotspot information with these databases to estimate the approximate location of the device. The Wi-Fi positioning function can accurately determine the specific location of the user at home or at work, for example, identifying whether the user is in the kitchen or bedroom at home.

[0027] In this embodiment, the specific process of screening out the ten previous user behavior data of the same time period that best matches the real-time user behavior data is as follows: A is compared with the corresponding ai for similarity, and the first similarity weight value V1 is obtained; B is compared with the corresponding bi for similarity, and the second similarity weight value V2 is obtained; P is matched with the corresponding pi, and the straight-line distance between the two is calculated and marked as L; the matching degree S=(V1*V2 / L)*100% between the previous user behavior data of the same time period and the real-time user behavior data is calculated; the ten previous user behavior data of the same time period with the largest S are screened out. The regularity of work and rest will make the user's daily activities very similar. By comparing and analyzing this similarity, it is possible to quickly determine where the user is currently doing and what he is doing. For example, the behavior data of the user when sleeping is usually not much different, but in the same time period, the user's current real-time behavior data is very similar to the previous behavior data when sleeping, and it is not difficult to analyze that the user is currently in the sleep stage.

[0028] In this embodiment, when the matching degree S is screened, if the ten largest matching degrees S screened out are all less than the preset matching degree, an update event is triggered, and the smart wearable watch notifies the user and the user imports the current scene information. If the similarity is very small, it means that the user's current action is different from the past, and the user is currently in a new scene. Therefore, the user needs to add this scene to the classifier.

[0029] In this embodiment, in the process of outputting the scene label with the most repetitions as the user's current scene, if there are multiple scene labels with the most repetitions, the matching degree between the previous user behavior data and the real-time user behavior data corresponding to the same time period of each scene label is extracted, and then the sum of the matching degrees corresponding to each scene label is calculated, and the scene label with the largest sum of matching degrees is output as the user's current scene (for example, there are currently two scene labels with the most repetitions and both are three times, and the matching degrees between the previous user behavior data and the real-time user behavior data corresponding to these two scene labels are also three, and the sum of the three matching degrees corresponding to the same scene label is the sum of the matching degrees corresponding to the scene label).

[0030] In this embodiment, the specific process of classifying the current wrist motion posture is as follows: obtaining the scene in which the user is currently located, that is, the corresponding scene label; comparing the captured wrist motion posture with the four preset wrist postures for similarity, if the similarity of any group of comparisons is greater than the preset value, proceeding to the next step, if the similarities of the four groups of comparisons are all less than or equal to the preset value, the current wrist motion posture is planned to the first sequence; extracting the scene label and judging whether the user is currently in the scene in which the do not disturb function is turned on according to the label, if the do not disturb function is turned on, the current wrist motion posture is planned to the second sequence, if the do not disturb function is not turned on, the current wrist motion posture is planned to the third sequence; if the user's current wrist motion posture is planned to the first sequence or the second sequence, the screen of the smart wearable watch is not awakened; if the user's current wrist motion posture is planned to the third sequence, the screen of the smart wearable watch is awakened. The fact that the current wrist motion posture is planned to the first sequence indicates that the user's posture does not trigger screen awakening, while the fact that the current wrist motion posture is planned to the second sequence indicates that the user's posture triggers screen awakening but is an inadvertent awakening under unexpected conditions.

[0031] In this embodiment, based on the user's operation on the smart wearable watch after the wrist motion posture is classified, the specific process of updating the state recognition model is as follows: if the current wrist motion posture is planned to the first sequence, and the user repeatedly wakes up or manually wakes up the smart wearable watch after the wrist motion posture is classified, the preset wrist posture is updated, that is, one of the four preset wrist postures is randomly replaced with the current user's wrist motion posture, wherein repeated wake-up means that the number of times the user raises his wrist exceeds the preset value within a preset time period; if the current wrist motion posture is planned to the second sequence, and the user repeatedly wakes up or manually wakes up the smart wearable watch after the wrist motion posture is classified, an update event is triggered, and the smart wearable watch notifies the user and the user imports the current scene information. This setting can update the preset wrist posture to keep up with the changes in the user's action habits, and can also update the scene information in the classifier to improve the accuracy of subsequent classification.

[0032] like Figure 3As shown, the present application also discloses an intelligent wear behavior classification system based on AI technology, which adopts an intelligent wear behavior classification method based on AI technology, including an intelligent wear watch and an AI analysis platform, wherein the AI ​​analysis platform includes a scene classification module, a scene analysis module, a posture classification module, and an update module; the intelligent wear watch is used to collect user behavior data and send the behavior data to the AI ​​analysis platform; wherein the behavior data includes real-time physiological sign data, real-time motion data, and positioning data; the scene classification module trains a pre-established classifier based on the collected previous user behavior data and the corresponding time to obtain a scene classification model; the scene analysis module analyzes the user's current scene based on the collected real-time user behavior data, combined with the current network time and the scene classification model; the posture classification module is used to capture the user's wrist motion posture, and classify the current wrist motion posture in combination with the user's current scene, and decide whether to wake up the screen of the intelligent wear watch according to the classification result; the update module decides whether to update the user's current scene based on the user's operation on the intelligent wear watch after the wrist motion posture classification.

[0033] The present invention can not only classify the user's current usage scenario in detail, but also analyze whether to wake up the screen of the smart wearable watch according to the classified usage scenario, so as to avoid waking up the smart watch in unnecessary scenarios to distract the user's attention or affect his rest, thereby avoiding causing trouble to the user and facilitating use.

[0034] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or replace some or all of the technical features therein by equivalents. These modifications or replacements do not make the essence of the corresponding technical solutions deviate from the scope of the technical solutions of the embodiments of the present invention, and they should all be included in the scope of the claims and specification of the present invention.

Claims

1. A smart wear behavior classification method based on AI technology, characterized in that: The following steps are involved: The smart wearable watch collects the user's behavior data and sends the behavior data to the AI ​​analysis platform; wherein the behavior data includes real-time physiological sign data, real-time motion data and positioning data; Based on the collected previous user behavior data and the corresponding time, the pre-established classifier is trained to obtain a scene classification model; Based on the collected real-time user behavior data, combined with the current network time and scene classification model, the user's current scene is analyzed; Capture the user's wrist motion posture, classify the current wrist motion posture based on the user's current scene, and decide whether to wake up the screen of the smart wearable watch based on the classification result; Based on the user's operation on the smart wearable watch after the wrist movement posture is classified, decide whether to update the user's current scene.

2. According to the AI ​​technology-based smart wear behavior classification method of claim 1, it is characterized in that: The specific training process of the scene classification model is as follows: Divide the time corresponding to the previous user behavior data into working days and weekends; The user behavior data belonging to working days are divided into working day clusters, and the user behavior data belonging to rest days are divided into rest day clusters; Divide a day into eight time periods, and organize the user behavior data in the workday cluster and the weekend cluster into corresponding sub-clusters according to the divided time periods; Import the scene information corresponding to the previous user behavior data to add scene tags to the user behavior data in each sub-cluster. The scene tags include the mark of the scene the user is in and whether the do not disturb function is turned on.

3. According to the AI ​​technology-based smart wear behavior classification method of claim 2, it is characterized in that: The specific analysis process of the user's current scene is as follows: Generate a real-time physiological sign radar chart A based on the collected real-time physiological sign data, wherein the physiological sign data includes heart rate, blood pressure, blood oxygen saturation and skin conductivity; Generate a real-time sports radar chart B based on the collected real-time sports data, wherein the real-time sports data includes step frequency, breathing frequency, calorie consumption and speed; Determine the user's current location P based on the collected positioning data; According to the current network time, determine the sub-cluster of the corresponding time period of the collected real-time user behavior data; Based on the user behavior data in the sub-cluster of the corresponding time period, generate the physiological sign radar chart ai and the movement radar chart bi of the same time period in the past, and the location pi, i=1···n, where n is a positive integer; Match the real-time physiological sign radar chart A, real-time motion radar chart B, and current location P corresponding to the real-time user behavior data with the groups of physiological sign radar charts ai, motion radar charts bi, and location pi in the same time period in the past one by one, and select the ten user behavior data in the same time period in the past that best match the real-time user behavior data; Extract the scene labels corresponding to the ten selected user behavior data in the same time period in the past; Based on the extracted scene tags, the scene tag with the most repetitions is output as the user's current scene.

4. According to the AI ​​technology-based smart wear behavior classification method of claim 3, it is characterized in that: The smart wearable watch supports a Wi-Fi positioning function. When determining the user's current location, if the user's current location P is the user's residence or workplace, the smart wearable watch further positions the user by identifying the Wi-Fi signal strength of the user's residence or workplace and combining it with the Wi-Fi location database.

5. According to the AI ​​technology-based smart wear behavior classification method of claim 4, it is characterized in that: The specific process of selecting the ten user behavior data in the same period of the past that best match the real-time user behavior data is as follows: Compare A with the corresponding ai to obtain a first similarity weight value V1; Compare B with the corresponding bi for similarity to obtain a second similarity weight value V2; Match P with the corresponding pi, calculate the straight-line distance between them and mark it as L; Calculate the matching degree S = (V1*V2 / L)*100% between the user behavior data of the same time period in the past and the real-time user behavior data; Filter out the ten user behavior data with the largest S in the same time period.

6. The AI-based smart wear behavior classification method according to claim 5, characterized in that: When the matching degree S is screened, if the ten largest matching degrees S screened out are all less than the preset matching degree, an update event is triggered, the smart wearable watch notifies the user and the user imports the current scene information.

7. The AI-based smart wear behavior classification method according to claim 6, characterized in that: In the process of outputting the scene label with the most repetitions as the user's current scene, if there are multiple scene labels with the most repetitions, the matching degree between the user behavior data of the same time period in the past and the real-time user behavior data corresponding to each scene label is extracted, and then the sum of the matching degrees corresponding to each scene label is calculated, and the scene label with the largest sum of matching degrees is output as the user's current scene.

8. The AI-based smart wear behavior classification method according to claim 7, characterized in that: The specific process of classifying the current wrist movement posture is as follows: Get the user's current scene, that is, the corresponding scene label; The captured wrist motion posture is compared with the four preset wrist postures for similarity. If the similarity of any group of comparisons is greater than a preset value, the next step is entered. If the similarities of the four groups of comparisons are all less than or equal to the preset value, the current wrist motion posture is planned into the first sequence. Extract the scene tag and determine whether the user's current scene is open for Do Not Disturb based on the tag. If Do Not Disturb is turned on, the current wrist movement posture is planned to the second sequence; if Do Not Disturb is not turned on, the current wrist movement posture is planned to the third sequence; If the user's current wrist motion posture is planned to the first sequence or the second sequence, the screen of the smart wearable watch will not be woken up; If the user's current wrist motion posture is planned to the third sequence, the screen of the smart wearable watch will be awakened.

9. The AI-based smart wear behavior classification method according to claim 8, characterized in that: The specific process of updating the state recognition model based on the user's operation on the smart wearable watch after the wrist movement posture classification is as follows: If the current wrist motion posture is planned to the first sequence, and the user repeatedly wakes up the smart wearable watch or manually wakes up the smart wearable watch after the wrist motion posture is classified, the preset wrist posture is updated, that is, one of the four preset wrist postures is randomly replaced with the current user's wrist motion posture, where repeated wake-up means that the number of times the user raises his wrist exceeds the preset value within the preset time period; If the current wrist motion posture is planned to the second sequence, and the user repeatedly wakes up or manually wakes up the smart wearable watch after the wrist motion posture is classified, an update event is triggered, the smart wearable watch notifies the user, and the user imports the current scene information.

10. An AI-based smart wear behavior classification system, using the AI-based smart wear behavior classification method according to any one of claims 1 to 9, characterized in that: It includes a smart wearable watch and an AI analysis platform, wherein the AI ​​analysis platform includes a scene classification module, a scene analysis module, a posture classification module, and an update module; The smart wearable watch is used to collect user behavior data and send the behavior data to the AI ​​analysis platform; wherein the behavior data includes real-time physiological sign data, real-time motion data and positioning data; The scene classification module trains a pre-established classifier based on the collected previous user behavior data and the corresponding time to obtain a scene classification model; The scene analysis module analyzes the user's current scene based on the collected real-time user behavior data, combined with the current network time and scene classification model; The posture classification module is used to capture the user's wrist motion posture, classify the current wrist motion posture in combination with the user's current scene, and decide whether to wake up the screen of the smart wearable watch according to the classification result; The update module determines whether to update the user's current scene based on the user's operation on the smart wearable watch after the wrist movement posture is classified.

Citation Information

Patent Citations

  • Semi-supervised scene classification identification method, system and device based on multimode fusion

    CN109697453A

  • Behavior classification method and system based on intelligent wearable device

    CN109961090A

  • Display control method and wearable equipment

    CN113552937A

  • Interface switching method and device, wearable electronic equipment and storage medium

    CN114223139A

  • Heart rate detection method based on deep learning and wearable device

    CN115702782A