A long-term body posture monitoring and intelligent reminding system based on multi-body segment fusion

Abstract

This invention discloses a long-term posture monitoring and intelligent reminder system based on multi-segment fusion, comprising a wearable sensing module, a data processing and main control module, and a user terminal application module. The wearable sensing module collects motion and posture data of the head, neck, chest, pelvis, and lower limbs through inertial measurement units and flexible bending sensors deployed at multiple key parts of the user's body. The data processing and main control module executes a two-level process: first, it identifies the activity scene based on pelvic and thigh data; then, under a defined scene, it fuses multi-sensor data for specific posture analysis. Its integrated individualized threshold adaptive unit can dynamically update the judgment threshold based on the user's baseline and long-term monitoring data. The system adopts a duration-based low-intrusion reminder strategy, triggering an instruction only when poor posture continuously exceeds the threshold corresponding to the type and the number of reminders per unit time has not reached the upper limit. This invention achieves comprehensive, accurate, and personalized posture monitoring and intelligent reminders.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of health monitoring, in particular to a long-term body posture monitoring and intelligent reminding system based on multi-segment fusion. BACKGROUND

[0002] With the change of modern work and life style, the problem of poor body posture such as head forward and humpback is increasingly common, which has become an important inducement to cause chronic musculoskeletal pain and dysfunction, and poses a continuous threat to individual health and quality of life. There is a clear demand for developing a technical solution that can monitor, evaluate and improve body posture for a long time.

[0003] However, the existing body posture monitoring products and technologies still have obvious limitations. In terms of comprehensive monitoring, most solutions only target a single part (such as only monitoring the back or neck), and cannot reflect the coordinated posture changes of the head, neck, chest, pelvis and lower limbs in different scenarios such as sitting, standing and walking, resulting in one-sided evaluation results. In terms of scene adaptability, the system usually cannot identify whether the user is currently in a sitting, standing or walking state, resulting in high false alarm rate due to misjudgment of normal transient action as poor posture. In terms of judgment standard, a fixed threshold based on group data is generally used, which is difficult to adapt to the differences in physiological structure and habitual posture of individual users, and the monitoring accuracy is insufficient. In terms of reminding mechanism, it mostly relies on instantaneous over-standard triggering reminder, and lacks judgment on the duration of poor posture, causing frequent interference when the user reasonably leans forward, resulting in poor experience and low compliance. In terms of system implementation, the existing multi-point monitoring solutions are often inconvenient to wear, have low integration, and complex data processing, which is difficult to meet the daily long-term wearing demand. SUMMARY

[0004] Therefore, the embodiment of the present application provides a long-term body posture monitoring and intelligent reminding system based on multi-segment fusion to solve or improve the technical problems in the prior art.

[0005] The technical solution of the embodiment of the present application is as follows: a long-term body posture monitoring and intelligent reminding system based on multi-segment fusion, comprising: a wearable sensing module, including a chest-mounted vest, a smart waistband and left and right thigh straps, for collecting motion and posture data of multiple segments of a user; a data processing and main control module integrated in the smart waistband, for receiving and fusion processing the data of the wearable sensing module, identifying the user's activity scene and specific body posture problems, and generating a reminding instruction according to a preset strategy; a user terminal application module for receiving data from the data processing and main control module, performing body posture analysis, report generation and providing interaction and improvement suggestions.

[0006] Compared with the prior art, the beneficial effects of the present application are: through the combination of multiple inertial measurement units deployed on the neck, chest, pelvis and thighs and flexible bending sensors, multi-segment fusion perception and collaborative analysis of the head, neck, chest, pelvis and lower limbs are realized, and the comprehensiveness and overall evaluation accuracy of the monitoring are significantly improved; through the establishment of individualized posture baseline and dynamic updating of threshold based on sliding window mean strategy, individualized adaptation of judgment standard is realized, and the precision and user adaptability of the monitoring are improved; through the two-level processing flow of first identifying the current activity scene such as sitting, standing and walking based on pelvis and thigh data, and then fusion analysis of the specific posture of the upper body in the scene, the misjudgment rate is effectively reduced, and the scene adaptability and robustness of the system are enhanced; by introducing the judgment logic based on the duration of different body state problems and setting the upper limit of the number of reminders per unit time, intelligent reminding with low disturbance is realized, which significantly improves the user experience and long-term compliance while ensuring effective intervention. BRIEF DESCRIPTION OF DRAWINGS

[0007] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, a brief introduction will be given below to the drawings needed to be used in the embodiments or prior art description. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative effort on the basis of these drawings.

[0008] Figure 1 is the system overall framework diagram of the present application; Figure 2 is the two-level data processing flowchart of the present application; Figure 3 is the individualized threshold self-adaptive unit working flowchart of the present application; Figure 4 is the low-disturbance reminding strategy logic diagram of the present application. DETAILED DESCRIPTION

[0009] In the following, only some exemplary embodiments are simply described. As those skilled in the art can recognize, the described embodiments can be modified in various different ways without departing from the spirit or scope of the present application. Therefore, the drawings and description are considered to be exemplary in nature rather than limiting.

[0010] The embodiments of the present application will be described in detail below with reference to the drawings.

[0011] The present application proposes a long-term body state monitoring and intelligent reminding system based on multi-segment fusion. The system comprises: A wearable sensing module, including a shirt, a smart waistband and left and right thigh straps, for collecting motion and posture data of multiple segments of the user; The data processing and main control module is integrated into the smart belt. It is used to receive and process the data from the wearable sensing module, identify user activity scenarios and specific postural problems, and generate reminder instructions according to preset strategies. The user terminal application module is used to receive data from the data processing and main control module, perform posture analysis, generate reports, and provide interactive and improvement suggestions.

[0012] This system achieves comprehensive, accurate, and low-interference long-term monitoring and intelligent reminders of users' body posture through multi-segment data acquisition, intelligent scene recognition, and individualized threshold processing, effectively improving user experience and compliance.

[0013] It is worth noting that wearable sensing modules refer to a collection of sensors that come into direct contact with or are worn on the user's body, used to collect kinematic and postural data of different parts of the user's body in real time and continuously. This module is integrated into clothing or straps in a non-invasive manner to ensure wearing comfort and long-term usability.

[0014] The data processing and main control module is responsible for receiving raw data from the wearable sensing module and preprocessing, fusing, analyzing, and interpreting it. This module can identify the user's activity scenarios, determine if there are specific postural problems, and generate corresponding reminder instructions based on preset strategies. This module has a certain level of computing and storage capabilities and can communicate with the user terminal application module.

[0015] The user terminal application module exists as an application on smartphones, tablets, or other smart devices. It receives and displays posture analysis results and reports sent by the data processing and main control module. This module also provides a user interface, allowing users to view historical data, set personalized parameters, and receive posture improvement suggestions and correction guidance provided by the system.

[0016] Multi-segment fusion refers to a system that simultaneously collects and comprehensively analyzes motion and posture data from multiple key parts of a user's body (such as the neck, chest, pelvis, and thighs) to gain a comprehensive and three-dimensional understanding of the user's overall posture. This fusion process can more accurately identify complex postural problems and coordination anomalies between body segments.

[0017] Long-term posture monitoring refers to the system's ability to continuously and uninterruptedly track and record a user's posture. Through long-term monitoring, the system can accumulate a large amount of posture data, thereby analyzing trends in posture changes, identifying common posture problems, and providing users with personalized health management suggestions based on time series data.

[0018] An intelligent reminder system refers to a system that, upon detecting poor posture in a user, can issue a low-intrusion reminder based on a preset intelligent strategy. This reminder mechanism considers the duration and severity of the posture problem, as well as the user's activity context, to provide timely feedback that does not disrupt the user's normal activities, thereby encouraging the user to proactively adjust their posture.

[0019] This embodiment provides a long-term posture monitoring and intelligent reminder system based on multi-body segment fusion, which mainly includes a wearable sensing module, a data processing and main control module, and a user terminal application module.

[0020] Wearable sensing modules are configured to collect motion and posture data from multiple body segments of the user. One implementation involves multiple independent sensor units, such as accelerometers and gyroscopes placed on the user's neck, chest, waist, and left and right thighs, respectively. These sensor units are connected to a central data acquisition unit via wires or wirelessly transmit data to a data processing and main control module. These sensors are simply fixed to the surface of the user's clothing or secured with elastic bandages. Therefore, raw motion data for each body segment, such as angular velocity and linear acceleration, can be acquired.

[0021] The data processing and main control module is configured to receive and fuse data from the wearable sensor module, identify user activity scenarios and specific postural problems, and generate reminder commands according to preset strategies. This module is a microcontroller unit integrated within the smart belt. The microcontroller receives raw data streams from various sensors and performs basic filtering. For example, by performing simple threshold checks on acceleration data, it can roughly distinguish whether the user is stationary or in motion. For postural problem identification, a set of fixed posture angle thresholds is used; for example, when the pitch angle detected by the neck sensor exceeds a certain fixed angle, it is determined to be forward head posture. Reminder command generation is based on instantaneous judgment; once poor posture is detected, a simple vibration or sound signal is triggered.

[0022] The user terminal application module is configured to receive data from the data processing and main control module, perform posture analysis, generate reports, and provide interactive and improvement suggestions. This module is an application running on a smartphone. The application receives and displays raw or pre-processed data sent by the data processing and main control module, for example, displaying the current tilt angle of the neck or waist in simple numerical form. Users can view real-time posture data and receive system alerts through the application. In addition, the application can provide basic historical data recording functions, such as recording the number of times poor posture occurs each day, but it does not perform in-depth statistical analysis of the data or generate complex reports.

[0023] The system proposed in this invention collects data from multiple body segments of the user through a wearable sensing module, overcoming the limitations of traditional solutions that rely on a single monitoring point, and achieving comprehensive perception of the user's overall posture. The data processing and main control module can fuse data from multiple body segments, identify activity scenarios and postural issues, and generate reminders, thereby improving the accuracy of posture recognition. The user terminal application module provides posture analysis, report generation, and improvement suggestions, offering users convenient interaction and personalized guidance. Therefore, this system provides users with a comprehensive, intelligent, and easily accessible long-term posture monitoring and intelligent reminder solution.

[0024] A long-term posture monitoring and intelligent reminder system based on multi-segment fusion uses a wearable sensing module to collect motion and posture data of multiple body segments. However, if the sensing module only uses general-purpose or limited-number sensors, it is difficult to comprehensively and accurately capture the complex multi-segment motion and posture information of the human body. Especially when it is necessary to analyze the relative posture of key body segments such as the neck, thoracic spine, pelvis, and thighs in detail, the accuracy and dimensionality of data acquisition will be limited, thus affecting the accuracy of subsequent posture recognition and reminders.

[0025] To address this, the present invention further proposes a specific configuration for a wearable sensing module, which includes a first six-axis inertial measurement unit (IMU) disposed around the neck of a garment close to the body, for acquiring the head's attitude quaternions. and acceleration data; a second IMU, positioned close to the chest of a shirt, provides a spatial reference for thoracic spine posture, and its output posture quaternion is... A third IMU integrated into the smart belt is used to acquire pelvic posture quaternions. and angle data; the fourth and fifth IMUs, respectively deployed on the left and right thigh straps, are used to acquire the posture and motion data of the left and right thighs, and their posture quaternions are respectively and ; and a flexible bending sensor installed on the back of a close-fitting garment, arranged along the thoracic spine, whose resistance or voltage output is related to the deformation. ΔL The deformation is proportional to the change in shape. ΔL Used to calculate the thoracic vertebral curvature angle.

[0026] Specifically, the first six-axis inertial measurement unit (IMU) is positioned close to the neck of the garment, and its main function is to acquire the user's head attitude quaternion. And acceleration data. A six-axis inertial measurement unit typically integrates a three-axis accelerometer and a three-axis gyroscope, capable of measuring the linear acceleration and angular velocity of an object in three-dimensional space. By performing attitude calculations on these raw data, the spatial attitude of the head can be calculated in real time, expressed as quaternions. The data is represented in a numerical form, while also providing acceleration information for head movements. This is crucial for assessing poor postures such as forward head posture and head yaw. The second IMU is positioned close to the chest area of ​​the shirt and serves as a spatial reference for thoracic spine posture. Similar to the first six-axis inertial measurement unit (IMU), the second IMU also measures chest motion data using its internal accelerometers and gyroscopes, and calculates the chest posture quaternion. The chest, as the core of the torso, provides posture data that can serve as a relative reference point for posture analysis of other body segments (such as the neck and pelvis), aiding in the creation of accurate body posture models. A third IMU, integrated into the smart belt, is used to acquire pelvic posture quaternions. And angle data. The pelvis is the cornerstone of human posture, and its tilt, rotation, and other postural changes have a significant impact on spinal health and overall balance. The third IMU, by measuring motion data in the pelvic region and performing posture calculations, can provide precise posture information of the pelvis in three-dimensional space, including anterior tilt angle, lateral tilt angle, etc. This data is a key basis for assessing sitting posture, standing posture, and pelvic health. The fourth and fifth IMUs are respectively placed on the left and right thigh straps to acquire posture and motion data of the left and right thighs, and their posture quaternions are as follows. and By monitoring the posture of the left and right thighs separately, bad habits such as crossing legs and leg internal / external rotation can be identified, and gait symmetry can be assessed. These IMUs also measure the movement of the thighs by integrating accelerometers and gyroscopes, and calculate their respective posture quaternions, providing detailed data for lower limb motion analysis. Flexible bending sensors are installed on the back of a close-fitting garment, arranged along the thoracic spine. These sensors are based on the principle of resistance or capacitance; when they undergo bending deformation, their resistance or voltage output changes accordingly, and this change is related to the deformation. ΔL The curves are proportional. By measuring this output change, the curvature angle of the thoracic spine can be calculated in real time. Flexible curvature sensors can directly reflect the degree of curvature of the thoracic spine. When used in conjunction with an inertial measurement unit (IMU), they can more accurately assess problems such as kyphosis and scoliosis, providing more intuitive deformation data.

[0027] By deploying multiple inertial measurement units (IMUs) and flexible bending sensors at specific locations within a wearable sensing module, this invention enables comprehensive and detailed acquisition of motion and posture data for multiple key body segments of the user (including the neck, chest, pelvis, and left and right thighs). The first six-axis IMU and the second IMU capture the posture of the head and chest, respectively, providing a foundation for upper torso posture analysis; the third IMU focuses on pelvic posture, providing crucial data for lower torso stability and tilt; the fourth and fifth IMUs record the motion and posture of the left and right thighs in detail, helping to identify undesirable lower limb habits. Furthermore, the flexible bending sensors, arranged along the thoracic spine, can directly quantify the degree of thoracic spine curvature, overcoming the limitations of a single inertial measurement unit (IMU) in deformation measurement. This fusion configuration of multiple body segments and types of sensors significantly improves the dimensionality and accuracy of data acquisition, enabling the system to more accurately identify specific postural problems of users in different activity scenarios, such as forward head posture, thoracic lordosis, pelvic tilt, and crossing legs. This provides a reliable and rich data foundation for subsequent data processing and intelligent reminders, thereby effectively solving the limitation of relying solely on general-purpose sensor modules to comprehensively capture complex human posture information.

[0028] In some embodiments of the present invention, a method is proposed to collect motion and posture data of multiple body segments of the user through a wearable sensing module, and then perform data processing and fusion processing by a main control module to identify user activity scenarios and specific postural problems. However, in practical applications, due to significant differences in body shape, habitual posture, and perception of "good" posture among individuals, using a uniform fixed threshold to judge poor posture may lead to frequent false alarms or missed alarms, thereby reducing user trust and compliance with the system and affecting the effectiveness of long-term monitoring and reminders.

[0029] Therefore, the present invention further proposes a data processing and main control module including an individualized threshold adaptive unit, which is configured to perform the following steps: (1) During the system initialization phase, an individual posture baseline vector is established based on the readings of each sensor when the user is in a standard standing posture:

[0030] In the formula, The initial pitch angle of the neck. The initial flexion angle of the thoracic vertebrae. This is the initial anterior pelvic tilt angle; (2) Based on the baseline vector Compared with the preset health reference range Calculate the initial individualization threshold range:

[0031] In the formula, This refers to the individual tolerance bias obtained based on statistical learning. (3) During long-term monitoring, based on the user's actual posture data The dynamic distribution is optimized using a sliding window mean update strategy, iteratively optimizing the threshold interval. The specific update formula is as follows:

[0032] In the formula The forgetting factor (value range 0.8–0.95). It serves as the statistical center for user posture data within the most recent time window.

[0033] The individualized threshold adaptive unit is a core component of the data processing and main control module. It dynamically adjusts the threshold used to judge poor posture based on the user's individual characteristics and long-term posture data, achieving more accurate and personalized posture monitoring. This ensures the system can adapt to the physiological differences and posture habits of different users, avoiding the inaccuracies that may arise from using a uniform standard.

[0034] During the system initialization phase, an individual posture baseline vector is established based on the readings of various sensors when the user is in a standard standing posture.

[0035] In the formula, The initial pitch angle of the neck. The initial flexion angle of the thoracic vertebrae. This is the initial anterior pelvic tilt angle.

[0036] Upon first use of the system, the user is guided to maintain a standard standing posture. At this time, the first six-axis inertial measurement unit (IMU), the flexible bending sensor, and the third IMU in the wearable sensing module will collect the corresponding posture data. The data processing and main control module uses this data to calculate the user's initial neck pitch angle in the standard standing posture. Initial flexion angle of the thoracic spine and initial anterior pelvic tilt angle These initial angles together constitute the individual pose baseline vector. This serves as a personalized reference point for subsequent posture assessments of the user. This establishes a unique, physiologically consistent normal posture benchmark for each user, making subsequent posture assessments more targeted.

[0037] Based on the baseline vector Compared with the preset health reference range Calculate the initial individualization threshold range ,inΔ This refers to the individual tolerance offset obtained based on statistical learning. This is after establishing the user's individual pose baseline vector. Then, the system will combine the preset health reference ranges. This health reference range is derived from general health standards in ergonomics, medicine, or sports science. Simultaneously, the system will incorporate individual tolerance biases derived from statistical learning. Δ .this Δ The value reflects the reasonable range of fluctuation in an individual's posture relative to an ideal baseline within a healthy range. (This is achieved by...) and Δ By combining these factors, the system can calculate the initial individualized threshold range. This range takes into account both the user's personalized baseline and general health standards as well as the natural fluctuations in individual posture, providing a preliminary basis for subsequent real-time monitoring.

[0038] During long-term monitoring, based on the user's actual posture data The dynamic distribution is optimized using a sliding window mean update strategy, iteratively optimizing the threshold interval. The specific update formula is as follows:

[0039] In the formula, α The forgetting factor (value range 0.8–0.95). It serves as the statistical center for user posture data within the most recent time window.

[0040] As the system monitors users over a long period, the user's actual posture data... Data will be continuously collected. To ensure that the threshold can adapt to long-term changes in user posture (e.g., posture improvement through corrective training, or posture drift due to changes in lifestyle), the system uses a sliding window mean update strategy to iteratively optimize the threshold range. Specifically, the system will maintain a sliding window containing the user's pose data over a recent period of time (e.g., several hours or days). This represents the statistical center (e.g., mean or median) of user pose data within that time window, reflecting the user's typical current pose. This is achieved through updating the formula. ,in It is a forgetting factor (typically between 0.8 and 0.95), which allows the system to update the threshold in a smooth and gradual manner. Forgetting factor This makes the new threshold The old threshold was largely retained. The information, while also appropriately incorporating This represents the latest posture trend. This iterative update mechanism ensures that the threshold is always synchronized with the user's actual posture, improving the accuracy and adaptability of monitoring.

[0041] Through the above technical solution, this system can effectively solve the problems of false alarms and missed alarms caused by the use of fixed thresholds in traditional posture monitoring systems. Firstly, in the system initialization phase, an individual posture baseline vector is established based on the user's standard standing posture. The system customizes a unique normal posture reference point for each user, fully considering the physiological differences and habitual postures of different individuals, making subsequent posture assessments more personalized and accurate. Secondly, based on the individual's baseline vector... Compared with the preset health reference interval and the individual tolerance bias obtained by statistical learning Calculate the initial individualization threshold range This not only ensures that the thresholds meet general health standards but also provides users with a reasonable and acceptable range of postures, avoiding overly strict or lenient judgments. Furthermore, during long-term monitoring, the system employs a sliding window mean update strategy, iteratively optimizing the threshold range based on the dynamic distribution of the user's actual posture data. This dynamic adaptation mechanism allows the threshold to smoothly adjust as user posture changes over time (e.g., improvements through corrective training or drift due to lifestyle habits), ensuring the continued relevance and effectiveness of alerts. For example, when a user improves poor posture through effort, the system can recognize this progress and adjust the threshold accordingly, avoiding unnecessary alerts due to old thresholds, thereby enhancing user compliance and motivation. Conversely, if a user's posture deteriorates over a long period, the threshold can also be adjusted to ensure timely detection and alerts. Therefore, this individualized threshold adaptation scheme significantly improves the accuracy of posture monitoring, user experience, and long-term intervention effects, enabling the system to truly become an effective tool for users to improve their posture.

[0042] In some embodiments of the present invention, a method is proposed to collect motion and posture data of multiple body segments of the user through a wearable sensing module, and to fuse this data with the main control module through data processing, thereby establishing a user-specific posture baseline and dynamic threshold. However, in actual long-term posture monitoring, users' activity scenarios are varied, such as sitting, standing, and walking. The evaluation criteria and focus of posture should differ in different scenarios. If posture analysis is performed directly without distinguishing between activity scenarios, it may lead to misjudgments or inaccurate alerts, reducing the system's usability and user experience.

[0043] In response, this invention further proposes that the data processing and main control module be configured to execute a two-level data processing flow: The first level is scene recognition: Based on the data from the third IMU (pelvis) and the fourth and fifth IMUs (thighs), the user's current activity scene is identified by analyzing tilt angle, relative angle and movement cycle characteristics. The activity scene includes sitting posture, standing posture, walking and posture transition state. The second level is posture analysis: Under the current activity scenario determined in the first level, data from the first IMU (neck), the second IMU (chest), the flexible bending sensor, and the third IMU (pelvis) are integrated to analyze and calculate specific posture problems.

[0044] Specifically, this two-stage data processing workflow breaks down the complex task of posture monitoring into two more manageable and accurate stages. The first stage focuses on macroscopic user behavior pattern recognition, i.e., identifying the user's current activity, which provides crucial contextual information for subsequent refined posture analysis. The second stage, based on the known activity scenario, performs a deeper and more accurate assessment of the posture of specific body parts. This hierarchical processing approach effectively improves the accuracy and robustness of posture recognition, avoiding inappropriate posture assessments in unsuitable scenarios.

[0045] The first-level scene recognition is a prerequisite for posture monitoring, providing accurate contextual information for subsequent posture analysis. For example, assessing forward head posture while seated and standing may involve completely different reference standards and focuses. This recognition process is based on data from the third IMU (pelvis) and the fourth and fifth IMUs (thighs). The third IMU (pelvis) provides posture information of the lower torso, such as the tilt angle of the pelvis, which is crucial for distinguishing between seated and standing postures. The fourth and fifth IMUs (thighs) provide data on thigh movement and posture, such as the change in the thigh's angle relative to the torso and the angular velocity of the thigh. This data is decisive for identifying walking states or distinguishing leg postures (such as crossing legs) between seated and standing postures. By analyzing the tilt angles, relative angles (such as the relative angle between the thigh and pelvis), and motion cycle characteristics (such as periodic swaying during walking) of these sensors, the system can accurately determine whether the user is currently in a seated, standing, or walking posture, or in a transitional state between these states. For example, sitting and standing postures can be distinguished by the pelvic tilt angle and the relative angle between the thigh and the trunk; walking can be identified by the periodic changes in thigh angular velocity.

[0046] After the first-level scene recognition determines the user's current activity scenario, the second-level posture analysis performs a more refined and accurate assessment of postural issues specific to that scenario. For example, after identifying a sitting scenario, the system will focus on analyzing issues such as forward head posture, thoracic lordosis, anterior pelvic tilt, and whether the user is crossing their legs. This analysis integrates data from the first IMU (neck), second IMU (chest), flexible bending sensor, and third IMU (pelvis). The first IMU (neck) provides head posture data for calculating the forward head posture angle. The second IMU (chest) and flexible bending sensor work together to accurately measure the degree of thoracic lordosis. The third IMU (pelvis) assesses pelvic tilt. By comprehensively utilizing this multi-segment sensor data in specific scenarios, the system can avoid cross-scenario misjudgments and specifically calculate and identify particular postural problems, such as forward head posture, excessive thoracic lordosis, anterior or posterior pelvic tilt, and unbalanced sitting postures (such as crossing legs).

[0047] When the scene is identified as a walking scenario, the posture analysis also includes calculating the head pitch angle based on the first IMU (neck) data, which is used to count the cumulative duration of the user walking with their head down.

[0048] Through the above technical solution, this invention decomposes the posture monitoring task into two stages: scene recognition and posture analysis, effectively solving the problem of potential misjudgment in posture assessment under varying activity scenarios. First, through the first-level scene recognition, the system can accurately determine the user's current activity state based on pelvic and thigh movement data, providing necessary contextual information for subsequent refined analysis. This preprocessing mechanism avoids applying a uniform posture assessment standard in inappropriate scenarios, thus significantly improving the accuracy of posture recognition. Second, after determining the activity scenario, the second-level posture analysis can specifically integrate data from neck, chest, pelvis, and flexible bending sensors to calculate and identify posture problems in specific scenarios. For example, in a sitting position, the system will focus more on assessing the static posture of the neck, thoracic spine, and pelvis, while in a walking scenario, it may focus more on gait balance. This hierarchical processing approach makes posture analysis more accurate and targeted, avoiding invalid or erroneous alerts, thereby improving the system's usability and user experience, and ensuring the effectiveness and reliability of long-term posture monitoring.

[0049] In some embodiments of the present invention, a long-term posture monitoring and intelligent reminder system based on multi-segment fusion is proposed. Its data processing and main control module is configured to execute a two-level data processing flow. The first level is scene recognition, used to identify the user's current activity scene. However, in practical applications, without specific and robust discrimination conditions, scene recognition based solely on sensor data may lead to misjudgments, especially when distinguishing common activity scenes such as sitting, standing, and walking. Confusion may arise due to individual differences or subtle movements, thus affecting the accuracy of subsequent posture analysis and the effectiveness of reminders.

[0050] In response, the present invention further proposes that the first-level scene recognition specifically includes: distinguishing between sitting and standing postures through specific discrimination conditions, and judging walking status through specific features.

[0051] Specifically, to distinguish between sitting and standing postures, the system uses the following criteria: when the anterior pelvic tilt angle... Greater than the sitting posture judgment threshold And the relative angle between the thigh and the torso Theoretical angle of the thigh when standing When the difference between the two values ​​is less than the allowable error ϵ, the system determines that the current scene is a sitting posture. Among them, the anterior pelvic tilt angle It is a key parameter for measuring the degree of pelvic tilt relative to the vertical direction. In a sitting position, the pelvis will exhibit a greater forward tilt angle than in a standing position. Sitting posture judgment threshold. This is a preset empirical value, usually set between 45° and 60°, used to define the critical point at which anterior pelvic tilt reaches the characteristics of a sitting posture. The relative angle between the thigh and torso. This reflects the posture of the thigh relative to the torso; this angle decreases significantly when sitting. Theoretical thigh angle when standing. This provides a reference standard for the standing position. Tolerance The existence of this makes the discrimination condition somewhat tolerant in actual measurement, avoiding frequent scene switching due to minor fluctuations.

[0052] To determine walking status, the system uses the following feature: the standard deviation of thigh angular velocity. Greater than the exercise intensity threshold Furthermore, a fast Fourier transform was performed on the thigh angular velocity. Subsequently, when a significant peak is observed within the frequency range of 0.5 to 2.5 Hz, the system determines the current scene to be walking. Among them, the standard deviation of the thigh angular velocity. This is an indicator of the intensity of thigh movement. During walking, the thighs swing periodically, causing significant fluctuations in angular velocity and consequently increasing the standard deviation. Exercise intensity threshold. Used to distinguish meaningful walking motion from unconscious body swaying. Fast Fourier Transform This is used to analyze the frequency components of thigh movements. The frequency of human walking usually falls in the range of 0.5 to 2.55 Hz. By detecting the peak values ​​in this frequency range, periodic walking movements can be effectively identified.

[0053] Through the above technical solution, this invention provides a more accurate and robust scene recognition method. By introducing posture features such as pelvic tilt angle and the relative angle between the thigh and torso, and combining them with preset thresholds and tolerances, the system can effectively distinguish between sitting and standing postures, avoiding misjudgments that may occur based solely on data from a single sensor. Simultaneously, by analyzing the standard deviation and frequency characteristics of the thigh angular velocity, the system can accurately identify periodic walking activities, distinguishing them from static or non-periodic movements. This refined scene recognition capability ensures that subsequent posture analysis is based on the correct activity context, thereby improving the accuracy of posture monitoring and the effectiveness of intelligent alerts. For example, for poor posture in a sitting position (such as forward head posture), targeted analysis and alerts can only be performed after accurately identifying the sitting scene, avoiding false alarms when standing or walking. This greatly enhances the system's practicality and user experience, reducing unnecessary interference.

[0054] In some of the embodiments of the present invention described above, a long-term posture monitoring and intelligent reminder system based on multi-body segment fusion is proposed. Its data processing and main control module can execute a two-level data processing flow, wherein the second level is posture analysis. In practical applications, if there is a lack of detailed methods for accurately quantifying and identifying specific posture problems, the system will have difficulty accurately judging the specific poor posture of the user, thereby affecting the effectiveness of subsequent reminders and correction suggestions.

[0055] In response, this invention further proposes a second-level posture analysis that specifically includes: calculation of the neck forward tilt angle, calculation of the thoracic spine curvature angle, calculation of the pelvic forward tilt angle, and judgment of the behavior of crossing one's legs.

[0056] Specifically, the forward tilt angle of the neck is calculated using the following formula:

[0057] In the formula, The head pitch angle is obtained from the three-axis output of the first IMU accelerometer after attitude calculation.

[0058] This calculation method utilizes accelerometer data from a first six-axis inertial measurement unit (IMU) positioned near the neck of a garment close to the body. The accelerometer measures the device's acceleration in inertial space; when the device is stationary or in uniform motion, it primarily experiences gravitational acceleration. The calculation is then performed using the arctangent function. This allows us to calculate the angle of the head in the pitch direction. Here... It is the acceleration component after attitude calculation, ensuring the accuracy of the angle. That is, it eliminates the interference of motion acceleration and mainly reflects the component in the direction of gravity, thus obtaining the pitch angle of the head relative to the direction of gravity. Accurately calculating this angle is crucial for identifying poor postures such as those of people looking down at their phones.

[0059] The thoracic vertebral curvature angle is calculated by fusing the deformation of the flexible sensor with the attitude of the second IMU:

[0060] In the formula, This represents the change in length of the flexible sensor. For calibration coefficients, The thoracic spine reference angle is provided for the second IMU. A flexible bending sensor, positioned on the back of a close-fitting garment and arranged along the thoracic spine, directly measures the thoracic spine deformation ΔL; its resistance or voltage output is proportional to the deformation ΔL. This is achieved through calibration coefficients. This allows the deformation to be converted into a preliminary bending angle. Simultaneously, a second IMU positioned on the chest area of ​​a close-fitting garment provides a spatial reference for the thoracic spine posture, and its output posture quaternion... The absolute posture angles of the thoracic vertebrae can be calculated. By fusing the relative deformation measurement of the flexible sensor with the absolute attitude measurement of the IMU, the thoracic vertebral curvature angle can be calculated more accurately. This overcomes the problem that a single sensor may have drift or local deformation that prevents it from reflecting the overall posture. The thoracic vertebral curvature angle is used to quantify the degree of curvature of the thoracic spine and is a key parameter for assessing poor posture such as kyphosis.

[0061] Anterior pelvic tilt angle is calculated using data from a third IMU:

[0062] In the formula, This represents the component of gravitational acceleration in the belt coordinate system. The method utilizes accelerometer data from a third IMU integrated into the smart belt. The accelerometer senses gravitational acceleration when the pelvis is in a relatively static state. This is expressed using the arctangent function. ,in It is the component of gravitational acceleration in the lumbar belt coordinate system, and the pitch angle of the pelvis relative to the direction of gravity can be directly calculated, i.e., the anterior pelvic tilt angle. This method can reflect the tilt state of the pelvis in real time. The anterior pelvic tilt angle is an important indicator for assessing pelvic posture and is closely related to lumbar spine health and overall standing and sitting posture.

[0063] The criteria for judging the behavior of crossing one's legs are:

[0064] in, The abduction angles calculated by the fourth and fifth IMUs, respectively. The angle difference threshold, γ represents the pelvic tilt angle, and γ is the tilt threshold. This judgment condition comprehensively considers the relative posture of the left and right thighs and the pelvic tilt. The fourth and fifth IMUs, respectively positioned on the left and right thigh straps, can acquire the posture and motion data of the left and right thighs. The abduction angle of the left and right thighs can be obtained through posture calculation. When the difference between the two Exceeding the preset angle difference threshold This indicates a significant asymmetrical abduction of the legs. Simultaneously, the third IMU integrated into the smart belt can calculate the pelvic tilt angle. When the pelvic tilt angle exceeds the preset tilt detection threshold γ, pelvic tilt is further confirmed. Only when both conditions are met simultaneously is it determined that the user is crossing their legs, thus avoiding false positives and improving the accuracy of identification. Crossing one's legs is a common poor sitting posture, and maintaining it for a long time may lead to problems such as pelvic tilt and scoliosis.

[0065] Through the detailed posture analysis methods described above, this system can accurately quantify and identify various specific postural problems, such as forward head posture, thoracic lordosis, anterior pelvic tilt, and leg crossing. This allows the data processing and main control modules to extract clear and identifiable postural issues from fuzzy posture data, providing a solid data foundation and judgment basis for subsequent individualized threshold adaptation, scene recognition, and intelligent reminder strategies. Compared to merely performing rough posture classification, this solution can gain in-depth insights into the details of users' poor posture, thereby achieving more accurate problem localization and more effective intervention, significantly improving the practicality and user experience of the long-term posture monitoring and intelligent reminder system.

[0066] In some of the above implementations, the system can identify user activity scenarios and specific postural issues, and generate reminder instructions based on preset strategies. However, if the reminder strategy is too sensitive or frequent, it will cause users to receive reminders too often, thereby causing excessive interference to users, reducing user experience and compliance.

[0067] To address this issue, this invention proposes a low-intrusion reminder strategy for the data processing and main control module, based on duration. The execution logic of this reminder strategy is as follows: a reminder instruction is generated only when the duration of a certain undesirable posture exceeds a preset duration threshold corresponding to that posture type; furthermore, the system sets an upper limit on the number of reminders triggered per unit time to prevent excessive interference with the user.

[0068] The duration-based low-intrusion alert strategy involves the data processing and main control module detecting poor posture in a user but not immediately issuing an alert. Instead, it assesses the duration of the poor posture. Only when the poor posture persists for a preset duration will the system generate an alert command. This avoids frequent alerts for brief, momentary postural deviations, thereby reducing unnecessary interference and increasing user acceptance of the system.

[0069] Specifically, after identifying a user's poor posture, the data processing and main control module will activate an internal timing mechanism. This mechanism will continuously monitor the duration of the poor posture. The system has preset duration thresholds for different types of poor postures. For example, the threshold may be longer for slight forward head posture, while the threshold will be shorter for extreme poor postures that could cause acute injury. Only when the detected duration of the poor posture exceeds its corresponding preset duration threshold will the data processing and main control module determine that intervention is needed and generate a corresponding reminder instruction.

[0070] Furthermore, to further optimize the user experience and prevent excessively frequent reminders due to repeated occurrences of poor posture or alternating poor postures within a short period, the data processing and control module has set an upper limit on the number of reminders triggered per unit of time. Even if a user repeatedly exhibits poor postures for durations exceeding a threshold within a certain time period, the system will control the reminder sending frequency based on the preset upper limit. For example, if the reminder count has reached the upper limit within a preset time window, subsequent reminders meeting the conditions will be temporarily suppressed until the next time window begins or the counter is reset. This is to avoid reminder fatigue and ensure the effectiveness of reminders and long-term user compliance.

[0071] By employing the aforementioned duration-based low-interference reminder strategy, this invention effectively addresses the potential for excessive interference in traditional reminder systems. Specifically, the system no longer immediately alerts users to fleeting or momentary postural instability. Instead, it determines whether the duration of the instability exceeds a preset threshold, ensuring that an alert is only triggered when the instability is truly persistent and potentially impacts health. Furthermore, by setting an upper limit on the number of alerts triggered per unit of time, this invention further avoids "reminder bombardment" caused by repeated occurrences or overlapping of multiple instability postures within a short period, significantly reducing user disruption. This strategy not only enhances user acceptance and compliance with reminders, making users more willing to use the system for long-term posture monitoring and improvement, but also focuses reminders on persistent postural instability that truly requires intervention, thereby improving the effectiveness and relevance of reminders and ultimately promoting healthier posture habits in users.

[0072] In some embodiments of the present invention described above, a duration-based low-interference reminder strategy is proposed. This strategy generates a reminder only when the duration of an unhealthy posture exceeds a preset threshold, and sets an upper limit on the number of reminders triggered per unit time to prevent excessive interference. However, in actual implementation, how to accurately determine the duration of the unhealthy posture and how to effectively manage the reminder triggering frequency per unit time to ensure the timeliness and accuracy of reminders without causing excessive user disturbance are technical details that require further clarification and optimization.

[0073] To address this, the present invention further proposes that the determination of the duration threshold be implemented through a timer, specifically: when the system detects an undesirable posture angle... Continuously exceeding the corresponding individualized dynamic threshold When the timer starts or accumulates, the timer continues until the preset duration threshold is reached. A reminder is triggered when the undesirable posture angle falls below a threshold during the timing period; if the angle falls below the threshold during the timing period, the timing is reset or paused. The maximum number of reminders per unit time is [not specified]. The determination is made through a unit time window. This is achieved by counting the number of times an alert has been triggered.

[0074] Specifically, the timer is a software timer or counter variable that runs within the data processing and main control module. When the system processes data collected by the wearable sensing module, it identifies the user's current undesirable posture angle. Continuously exceeding its corresponding individualized dynamic threshold When the timer is activated, it begins accumulating time. The timer increments its count at preset time intervals (e.g., per second) to accurately record the duration of the poor posture. The timer continues accumulating until it reaches a preset duration threshold. When the timer expires, the system will generate a corresponding reminder. To achieve a low-interference reminder strategy, if the user actively adjusts their posture during the timer period, correcting any undesirable posture angles... Falling back to the individualized dynamic threshold If the timer expires within a certain range, it will be immediately reset or paused. The reset operation is suitable when the user has completely corrected their posture, while the pause operation is suitable for scenarios where the user's posture has temporarily improved but there is still a risk of relapse. The timer can continue to accumulate when the poor posture reappears, thus avoiding unnecessary alerts.

[0075] In addition, to prevent the system from frequently triggering reminders in a short period of time and causing excessive interference to users, this invention also sets an upper limit on the number of reminders per unit time. The upper limit is determined by maintaining a reminder counter in the data processing and main control module. The system defines a unit time window. (For example, one hour), and count all successfully triggered reminder commands sent to the user terminal application module within this window. Before each reminder command is prepared to be triggered, the system checks the current unit time window. Has the number of reminders already triggered been reached? If the limit has been reached, this notification will be suppressed until the current time window. The end or counter is updated according to the sliding window mechanism, thereby effectively controlling the frequency of reminders.

[0076] Through the above technical solution, this invention can accurately determine the duration of poor posture, avoiding triggering alerts based solely on momentary judgments, thereby improving the accuracy and rationality of alerts. Simultaneously, by setting an upper limit on the number of alerts per unit time... And combined with time windows By using a counter, the system effectively avoids sending frequent reminders within a short period, significantly reducing user interference and improving the user experience. The timer's reset or pause mechanism ensures that users are not disturbed by unnecessary reminders after actively correcting their posture. This macroscopically controls the density of reminders, enabling the system to effectively remind users to correct poor posture while minimizing interference with their daily activities, achieving a good balance between the low-intrusion nature and effectiveness of the reminder strategy. Combined with individualized dynamic thresholds, the timer can make judgments based on the user's own posture habits and health status, making reminders more personalized and accurate, further improving user acceptance of the system.

[0077] In some embodiments of the present invention, a duration-based low-interference reminder strategy is proposed, which avoids excessive interference to users by setting a duration threshold for poor posture and an upper limit on the number of reminders per unit time. However, in practical applications, the degree of impact of different posture problems on physical health, their formation mechanisms, and users' perception of them all vary. If a uniform judgment condition and duration threshold are used, it may result in insufficient reminders for some minor but prolonged posture problems, or untimely reminders for some serious but short-lived posture problems, thereby affecting the effectiveness of the reminders and the user experience.

[0078] To address this, the present invention further proposes a reminder strategy that sets differentiated judgment conditions and duration thresholds for different types of postural problems, wherein the individualized dynamic thresholds used are dynamically generated according to the above method. Specifically: For forward head posture, the criterion is the forward head posture angle. Greater than the dynamic individualization threshold And continuous duration Greater than or equal to ,in The value range is 60–90 seconds; For thoracic curvature or pelvic tilt, the criteria for judgment are the thoracic curvature angle. Greater than the dynamic individualization threshold or anterior pelvic tilt angle Greater than the dynamic individualization threshold And satisfy the continuous duration Greater than or equal to ; The criterion for judging the behavior of crossing one's legs is that the legs are crossed in a seated setting. It is true and the state lasts for a continuous duration. Greater than or equal to ,in The range of values ​​is Second; For sustained sitting behavior, the judgment condition is the scenario. The continuous duration of sitting posture Greater than or equal to ,in The value range is 45–60 minutes; For exceeding the safety hard threshold set based on medical safety data Extremely poor posture, the criterion for judgment is the angle of poor posture. Greater than the safety hard threshold And continuous duration Greater than or equal to ,in The value range is 10–15 seconds, and it satisfies... Less than .

[0079] The implementation of the aforementioned reminder strategy relies primarily on the data processing and main control module's ability to identify different types of postural problems. This module can accurately determine the user's current activity scenario and specific postural problems, such as forward head posture, thoracic lordosis, pelvic tilt, crossing legs, or prolonged sitting, based on multi-segment data collected by the wearable sensing module through scene recognition and posture analysis.

[0080] Secondly, to achieve differentiated judgment conditions and duration thresholds, the data processing and main control module maintains a mapping table between posture problem types and corresponding alert rules. When a specific posture problem is identified, the system queries this mapping table to obtain the judgment logic and duration threshold for that posture problem. For example, for forward head posture, the system will monitor the acceleration data of the first six-axis inertial measurement unit (IMU) to calculate the forward head posture angle. And it is combined with the dynamic neck tilt threshold dynamically generated by the individualized threshold adaptive unit. A comparison is then made. Simultaneously, an internal timer records the duration for which the neck tilt angle consistently exceeds a threshold. Only when Persistently greater than and Achieve the preset The reminder will only be triggered after 60-90 seconds (for example).

[0081] Similarly, for thoracic flexion or pelvic tilt, the system monitors data from a flexible flexion sensor and a second IMU to calculate the thoracic flexion angle. And data from the third IMU to calculate the anterior pelvic tilt angle. These perspectives will be correlated with their respective dynamic individualized thresholds. or Compare and combine durations Has it been achieved? That determines whether to send a reminder.

[0082] Regarding the behavior of crossing one's legs, the system first uses scene recognition to determine whether the user is in a sitting position. If the person is seated, data from the fourth and fifth IMUs will be used to further determine whether the person is crossing their legs. When both conditions are met and the state continues for a certain duration... achieve The system will generate a reminder after 30-60 seconds (for example). Usually shorter than This reflects the asymmetrical pressure that crossing your legs can put on the pelvis and spine, and should be taken seriously even if the duration is relatively short.

[0083] For prolonged sitting behavior, the system only needs to determine the context. Whether one is in a seated position for an extended period of time, and the duration of such an extended period of time achieve The system will trigger a reminder at 45-60 minutes (for example) to encourage the user to get up and move around. It has a significantly longer threshold than other postural problems, which is used to remind users to get up and move around in time to avoid prolonged sitting.

[0084] In addition, to address extreme postural problems that could lead to acute injury, the system also sets a safety hard threshold. When any unfavorable posture angle When the hard threshold is exceeded, even if the duration is... Only (For example, 10-15 seconds), the system will also immediately trigger a reminder. much smaller This ensures rapid intervention in severe postures.

[0085] Through the above technical solution, this invention can achieve refined and personalized identification and alerts for different postural problems. This differentiated judgment criteria and duration thresholds, combined with individualized dynamic thresholds, make the alert strategy more accurate and effective. For example, for postures such as forward head posture and thoracic lordosis that require a long period of accumulation to manifest their harmful effects, a longer duration threshold is used. This avoids excessive reminders for brief, harmless postures, thus improving the user experience. For bad habits like crossing one's legs in specific scenarios, scene recognition and relatively short reminders are used. A threshold can be set to correct pelvic asymmetry in a timely manner, preventing its long-term accumulation. A longer threshold is set to address the common problem of prolonged sitting. Thresholds effectively remind users to move around when needed, preventing health risks associated with prolonged sitting. More importantly, a rapid response mechanism for extremely poor posture is introduced, utilizing medically safe hard thresholds. and extremely short The duration of the reminders ensures rapid intervention for serious postures that could lead to acute injury, greatly enhancing the system's safety and usability. This tiered and categorized reminder strategy avoids excessive intrusion while ensuring timely reminders for postural problems that truly require intervention, thereby significantly improving user compliance and postural improvement outcomes.

[0086] In some embodiments of the present invention described above, a long-term posture monitoring and intelligent reminder system based on multi-segment fusion is proposed. This system can effectively collect user posture data and identify poor posture, and generate reminder instructions according to preset strategies. However, in practical applications, users may need a deeper understanding of their posture problems, obtain personalized improvement plans, and track their improvement progress to maintain active participation and motivation, thereby achieving long-term behavioral changes and posture improvement.

[0087] In this regard, the present invention further proposes that the user terminal application module should provide at least the following functional units: data statistics and reporting unit, comprehensive scoring and trend unit, personalized correction guidance unit, and behavior attribution analysis unit.

[0088] The data statistics and reporting unit is used to visually display the types, cumulative duration, peak periods, and scenario-related analyses of poor posture on a daily / weekly basis. Specifically, this unit can present the poor posture information identified by the data processing and main control module to users in intuitive and easy-to-understand chart formats (such as bar charts, line charts, pie charts, heatmaps, etc.). For example, the system can generate a weekly report that clearly shows the frequency, average duration, and total cumulative duration of poor postures such as forward head posture, thoracic lordosis, and crossing legs over the past week. Simultaneously, this unit can analyze the distribution of these poor postures during peak periods (such as morning work hours and afternoon rest periods) and correlate them with specific activity scenarios (such as sitting and standing postures), helping users understand in which situations they are more prone to postural problems.

[0089] The comprehensive scoring and trend unit generates a radar chart of overall posture scores based on different activity scenarios and body parts, and plots a long-term posture improvement trend chart. Specifically, this unit quantitatively assesses a user's posture performance in different activity scenarios (such as sitting, standing, and walking) and in different body parts (such as the neck, thoracic spine, and pelvis), calculating corresponding health scores. These scores are visualized in radar chart format, allowing users to clearly see their performance strengths and weaknesses across various posture dimensions. Furthermore, this unit records and plots trend charts of these comprehensive scores over time, enabling users to intuitively track their long-term posture improvement or deterioration, thereby enhancing self-management awareness and motivation for continuous improvement.

[0090] The personalized correction guidance unit is used to push targeted correction training tutorials or immediate activity suggestions to users based on identified high-frequency or severe postural problems. Specifically, when the data processing and main control module identifies that a user has a high-frequency or severe postural problem, this unit can automatically or upon user request push customized correction training tutorials based on the specific type and severity of the problem. These tutorials can be videos or text-based instructions, guiding users to perform specific stretching, strengthening, or posture adjustment exercises. Simultaneously, this unit can also provide immediate activity suggestions, such as reminding users to get up and move around after prolonged sitting, perform simple stretches, or adjust their sitting posture to reduce body stress, thereby helping users proactively correct poor posture and develop healthy habits.

[0091] The behavioral attribution analysis unit is used to analyze the potential causes of postural problems based on long-term, multi-dimensional postural and behavioral data, and generate behavioral pattern analysis reports. Specifically, this unit can utilize data analysis and machine learning algorithms to deeply mine and analyze the user's long-term accumulated postural data, activity scenario data, and reminder feedback data. Through this analysis, the system can identify potential behavioral patterns or lifestyle habits that lead to poor posture. For example, the system may find that a user is more prone to thoracic lordosis during specific work tasks, or habitually crosses their legs while watching videos. Based on these analysis results, the system can generate behavioral pattern analysis reports to help users understand the root causes of their postural problems, thereby enabling intervention and improvement at the source.

[0092] By integrating a data statistics and reporting unit, a comprehensive scoring and trend unit, a personalized correction guidance unit, and a behavioral attribution analysis unit into the user terminal application module, this invention can transform raw posture monitoring data into feedback information that is easy for users to understand and operate. The data statistics and reporting unit provides a comprehensive overview of poor posture, enabling users to clearly understand the type, frequency, and occurrence scenarios of their posture problems. The comprehensive scoring and trend unit, through quantitative assessment and visualized trends, allows users to intuitively track the long-term progress of their posture improvement, thereby enhancing user participation and self-management awareness. The personalized correction guidance unit provides targeted correction plans and immediate suggestions, helping users proactively correct poor posture and develop healthy habits. The behavioral attribution analysis unit further explores the underlying causes of posture problems, enabling users to adjust the behavioral patterns that lead to poor posture, significantly improving user experience and long-term health benefits. I. Application Scenarios: A 28-year-old male office worker (user "Zhang San", height 175cm, weight 70kg) was selected as the monitoring subject. He works as a software engineer, spends more than 8 hours a day working at his desk, and has experienced mild neck and shoulder pain. He hopes to improve his posture, including forward head posture, rounded shoulders, hunchback, and crossing his legs. The system has already fitted Zhang San with sensor modules (a close-fitting shirt, a smart belt, and straps to his left and right thighs) and completed its initial calibration.

[0093] II. Specific Implementation and Data Calculation at Each Stage: (I) System initialization and individualized baseline establishment: Standard standing posture data acquisition: Guide Zhang San to stand against the wall, with his head, shoulders, and buttocks touching the wall, and his heels about 5cm away from the wall, maintaining natural breathing for 30 seconds. The system collects stable readings from each sensor during this period.

[0094] Individual attitude baseline vector calculation: Based on the initial data collected, Zhang San's individual posture baseline vector was calculated. :

[0095] Initial neck tilt angle (slight forward tilt, which is within the common range); Initial thoracic vertebral curvature angle (with slight kyphosis); Initial pelvic tilt angle (within the normal pelvic tilt range); Initial individualization threshold range generation: Preset health reference range For the neck [0°, 20°], thoracic spine [0°, 25°], and pelvis [0°, 15°].

[0096] Zhang San's individual tolerance offset obtained based on statistical learning .

[0097] Calculate the initial threshold interval :

[0098] (II) Long-term monitoring and dynamic threshold updates (taking the morning of the 3rd day as an example): The system has accumulated data after two days of monitoring. The threshold is updated using a sliding window (window spanning the past 24 hours, forgetting factor α=0.9).

[0099] Attitude data statistics center within the current window

[0100] previous threshold

[0101] updated threshold :

[0102] (Note: This is a simplified version of vector operations; in reality, the upper and lower bounds of the interval are updated separately.) (III) A complete process for identifying and triggering alerts regarding poor posture: Time context: Day 3 of monitoring, from 10:00 AM to 10:15 AM, Zhang San was focused on writing code.

[0103] S1: Real-time data acquisition and scene recognition for multiple body segments: Table 1: Real-time monitoring data stream (sampling period 1 second, key time points selected)

[0104] Scene recognition logic: At 10:00:00, the sitting posture discrimination condition is met, and the system locks the scene as Sitting.

[0105] At 10:05:30, an angle difference of more than 25° between the left and right thighs and a pelvic tilt of more than 5° were detected, indicating the presence of "crossing legs" behavior (CrossLeg=True).

[0106] 10:06:00, attitude adjustment, CrossLeg=False.

[0107] S2: Specific Posture Problem Analysis (in the Sitting Scenario): The system integrates sensor data from the neck, chest, and pelvis to perform posture calculations.

[0108] Table 2: Calculation of key parameters for attitude analysis (corresponding to the time points in Table 1)

[0109] Example of posture angle calculation (taking a 10:05:30 neck tilt angle as an example): First IMU accelerometer output (gravity component after attitude calculation):

[0110] calculate:

[0111] Posture problem assessment: From 10:05:30 onwards, the cervical lordosis angle and thoracic flexion angle continuously exceeded their respective dynamic individualized thresholds.

[0112] Although the anterior pelvic tilt angle increased, it did not exceed the threshold.

[0113] The "crossing legs" behavior occurred between 10:05:30 and 10:06:00, lasting for about 30 seconds.

[0114] 3. S3 & S4: Intelligent Reminder Strategy Judgment and Triggering The system makes decisions based on preset differentiated duration thresholds and unit time reminder limits.

[0115] Timer status: Neck tilt timer : Accumulation starts from 10:05:30.

[0116] Thoracic flexion timer : Accumulation starts from 10:05:30.

[0117] Leg-crossing timer : Starting at 10:05:30, it will be reset to zero at 10:06:00.

[0118] Continuous meditation timer : Start accumulating from 10:00:00.

[0119] Thresholds and Decisions: Preset duration threshold: cervical and thoracic vertebrae (seconds) Crossing your legs for seconds Sit quietly for 10 minutes .

[0120] Unit time window hours Maximum number of reminders No alerts were triggered in the window between 10:00 and 11:00 today.

[0121] Judgment process: 10:06:30: No alert will be triggered. Continue timing. Continue timing. Minutes much smaller T3 .

[0122] 10:06:45: First, the triggering conditions must be met. The system checks the number of reminders within a unit time window (0 times < 3 times), and immediately generates and sends a comprehensive reminder: "A hunchback posture has been detected for more than 1 minute. It is recommended to straighten your chest and back and retract your shoulder blades." After the reminder is triggered, and Reset. Unit window alert count +1.

[0123] (iv) Post-reminder effect verification and system learning: User response: Zhang San felt the belt vibrate and saw the phone notification, and immediately adjusted his posture: head up, chest out, shoulders back.

[0124] Data feedback: In the following minutes, monitoring data showed that: Falling back to , Falling back to All returned to within the individualization threshold.

[0125] Systematic learning: This persistent "thoracic flexion" event was recorded. In subsequent sliding window threshold updates, due to the prolonged duration and severity of this postural instability, The statistical center value of the mid-thoracic vertebral curvature angle may be slightly elevated, leading to The upper limit of the mid-thoracic spine threshold has been slightly increased (e.g., from 16.9° to 17.2°). This reflects the system's adaptive fine-tuning of user habitual postures, avoiding over-reminders of habits that have been improved but still persist.

[0126] Meanwhile, users' timely responses to reminders are recorded by the behavior attribution analysis unit and used to assess user compliance.

[0127] (v) Summary of the daily data report (example): Table 3: Summary of User Zhang San's Daily Body Posture Monitoring Report (Day 3) ; III. Conclusion: This embodiment fully demonstrates the entire process of the system, from initial calibration and long-term adaptive learning to real-time multi-segment data fusion and acquisition, accurate scene recognition, and quantitative analysis of specific posture problems. Finally, it triggers low-intrusion reminders through a duration-based intelligent decision strategy. The example shows that the system can effectively distinguish between brief movements and persistent bad habits, greatly improving user experience and compliance while ensuring monitoring accuracy. Furthermore, through long-term data accumulation, it provides users with valuable insights into posture improvement and personalized guidance.

[0128] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any person skilled in the art can easily conceive of various variations or substitutions within the technical scope disclosed in the present invention, and these should all be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A long-term posture monitoring and intelligent reminder system based on multi-segment fusion, characterized in that, include: Wearable sensing modules, including a close-fitting top, a smart belt, and left and right thigh straps, are used to collect motion and posture data of multiple body segments of the user; The data processing and main control module, integrated into the smart belt, is used to receive and process the data from the wearable sensing module, identify user activity scenarios and specific postural problems, and generate reminder instructions according to preset strategies. The user terminal application module is used to receive data from the data processing and main control module, perform posture analysis, generate reports, and provide interaction and improvement suggestions.

2. The long-term posture monitoring and intelligent reminder system based on multi-segment fusion according to claim 1, characterized in that, The wearable sensing module includes: A first six-axis inertial measurement unit (IMU) positioned at the neck of the close-fitting garment is used to acquire the head's attitude quaternion. and acceleration data; A second IMU, positioned on the chest area of ​​the close-fitting garment, provides a spatial reference for thoracic spine posture, and its output posture quaternion is... ; The third IMU integrated into the smart belt is used to acquire the pelvic posture quaternion. and angle data; The fourth and fifth IMUs, respectively positioned on the left and right thigh straps, are used to acquire the posture and motion data of the left and right thighs. Their posture quaternions are as follows: and ; A flexible bending sensor, positioned on the back of the close-fitting garment and arranged along the thoracic spine, outputs a resistance or voltage that correlates with the deformation. The deformation is proportional to the change in shape. Used to calculate the thoracic vertebral curvature angle.

3. The long-term posture monitoring and intelligent reminder system based on multi-segment fusion according to claim 2, characterized in that, The data processing and main control module includes an individualized threshold adaptive unit, which is configured to perform the following steps: (1) During the system initialization phase, an individual posture baseline vector is established based on the readings of each sensor when the user is in a standard standing posture: ； In the formula, The initial pitch angle of the neck. The initial flexion angle of the thoracic vertebrae. This is the initial anterior pelvic tilt angle; (2) Based on the baseline vector Compared with the preset health reference range Calculate the initial individualization threshold range: ； In the formula, This refers to the individual tolerance bias obtained based on statistical learning. (3) During long-term monitoring, based on the user's actual posture data The dynamic distribution is achieved by employing a sliding window mean update strategy to iteratively optimize the threshold interval. The specific update formula is as follows: ； In the formula The forgetting factor (value range 0.8–0.95). It serves as the statistical center for user posture data within the most recent time window.

4. The long-term posture monitoring and intelligent reminder system based on multi-segment fusion according to claim 3, characterized in that, The data processing and main control module is configured to execute a two-level data processing flow: The first level is scene recognition: Based on the data from the third IMU (pelvis) and the fourth and fifth IMUs (thighs), the user's current activity scene is identified by analyzing tilt angle, relative angle and movement cycle characteristics. The activity scene includes sitting posture, standing posture, walking and posture transition state. The second level is posture analysis: Under the current activity scenario determined in the first level, data from the first IMU (neck), the second IMU (chest), the flexible bending sensor, and the third IMU (pelvis) are integrated to analyze and calculate specific posture problems.

5. A long-term posture monitoring and intelligent reminder system based on multi-segment fusion according to claim 4, characterized in that, The first level of scene recognition specifically includes: The following criteria can be used to distinguish between sitting and standing postures: ； In the formula Anterior pelvic tilt angle, The threshold for determining sitting posture (generally 45°–60°). The angle between the thigh and the torso. The theoretical angle of the thigh when standing. Tolerance; The following characteristics can be used to determine walking status: ； In the formula The standard deviation of the thigh angular velocity. The threshold for exercise intensity. This represents the Fast Fourier Transform, used for step frequency detection.

6. The long-term posture monitoring and intelligent reminder system based on multi-segment fusion according to claim 4, characterized in that, The second level of attitude analysis specifically includes: The forward tilt angle of the neck is calculated using the following formula: ； in The head pitch angle is obtained from the three-axis output of the first IMU accelerometer after attitude calculation. The thoracic vertebral curvature angle is calculated by fusing the deformation of the flexible sensor with the attitude of the second IMU: ； In the formula This represents the change in length of the flexible sensor. For calibration coefficients, The thoracic vertebral reference angle provided for the second IMU; Anterior pelvic tilt angle is calculated using data from a third IMU: ； In the formula This represents the component of gravitational acceleration in the belt coordinate system; The criteria for judging the behavior of crossing one's legs are: ； in The abduction angles calculated by the fourth and fifth IMUs, respectively. The angle difference threshold, The pelvic tilt angle. The threshold for determining tilt.

7. The long-term posture monitoring and intelligent reminder system based on multi-segment fusion according to claim 1, characterized in that, The data processing and main control module's reminder strategy is a low-intrusion reminder based on duration, and its execution logic is as follows: A reminder instruction is generated only when the duration of a certain poor posture exceeds a preset duration threshold corresponding to that posture type. In addition, the system sets a maximum number of reminders that can be triggered per unit of time to prevent excessive interference with users.

8. A long-term posture monitoring and intelligent reminder system based on multi-segment fusion according to claim 7, characterized in that, The determination of the duration threshold is achieved through a timer, specifically: When the system detects an undesirable posture angle Continuously exceeding the corresponding individualized dynamic threshold When the timer starts or accumulates, the timer continues until the preset duration threshold is reached. When the timer is active, a reminder is triggered; if the undesirable posture angle falls back to within the threshold during the timer period, the timer is reset or paused. The maximum number of reminders per unit time The determination is made through a unit time window. This is achieved by counting the number of times an alert has been triggered.

9. A long-term posture monitoring and intelligent reminder system based on multi-segment fusion according to claim 8, characterized in that, The reminder strategy sets differentiated judgment conditions and duration thresholds for different types of postural problems, wherein the individualized dynamic thresholds used are... Dynamically generated according to the method described in claim 3: For forward head posture, the criteria for judgment are: And continuous duration ,in Second, For dynamic individualized thresholds; The criteria for judging thoracic lordosis or pelvic tilt are as follows: or And satisfy ; The criterion for judging the behavior of crossing one's legs is that it occurs in a sitting setting. And the duration of this state ,in Second; The criteria for judging sustained sitting behavior are: Continuous duration ,in minute; For exceeding the safety hard threshold set based on medical safety data Extremely poor body posture, the criteria for judgment are: And continuous duration ,in seconds, and satisfy .

10. A long-term posture monitoring and intelligent reminder system based on multi-segment fusion according to claim 1, characterized in that, The user terminal application module provides at least the following functional units: Data statistics and reporting unit: used to visually display the types, cumulative duration, high-incidence periods, and scene correlation analysis of poor physical posture on a daily / weekly basis; Comprehensive scoring and trend unit: used to generate a comprehensive posture scoring radar chart based on different activity scenarios and body parts, and to draw a long-term posture improvement trend chart; Personalized correction guidance unit: used to push targeted correction training tutorials or instant activity suggestions to users based on the identified high-frequency or severe postural problems; Behavioral attribution analysis unit: Used to analyze the potential causes of postural problems based on long-term, multi-dimensional postural and behavioral data, and generate behavioral pattern analysis reports.

Images