An intelligent sitting posture optimization cushion system and intervention method thereof

Abstract

This invention relates to the fields of health monitoring and human factors engineering, and discloses an intelligent posture optimization pad system and its intervention method, addressing the problems of existing posture intervention products such as single identification, rigid intervention, lack of systematic incentives, and lack of ecological integration. The system includes a core sensing and processing unit, an interaction and execution unit, and an enterprise-level management and ecological interface unit. It collects data through a multimodal pressure sensor array, and an edge computing module combined with a lightweight AI model achieves accurate posture recognition. Personalized intervention strategies are generated based on a COM-B model. The intervention method constructs a complete closed loop of "perception-intervention-incentive-conversion," achieving deep integration of personal posture management and enterprise health management ecosystems. This invention improves the compliance and long-term health benefits of posture improvement, realizing a paradigm shift from individual hardware to organizational solutions. The scope of protection covers the overall system architecture and each core module, the entire intervention method process, and key optimization strategies.

Description

Technical Field

[0001] This invention relates to the field of carbon emission technology, specifically to an intelligent posture optimization pad system based on multimodal sensing and behavioral feedback, and its intervention method. Background Technology

[0002] Belt conveyors are a continuous, fast, and efficient material handling system widely used in industries such as coal, power, building materials, chemicals, machinery, and light industry. However, a survey conducted by the U.S. Mining Technicians Association on 7.2 billion tons of mining operations worldwide found that in more than half of the transport disruptions, conveyor belt failures caused by idler roller malfunctions led to unexpected interruptions. Each such disruption requires at least four hours of repair time, resulting in economic losses of thousands of dollars per minute.

[0003] Traditional manual inspection suffers from low efficiency and high missed detection rates, making it difficult to detect early idler damage. Existing research generally uses inspection carts to replace manual labor. For example, patents CN221294984U and CN218706140U use inspection carts to scan the barcodes or QR codes on the idlers to determine their rotation status, achieving non-contact inspection. However, this method can only monitor idler jamming and cannot collect vibration data to assess early idler faults or health status. Patent CN117383200A uses distributed optical fiber to monitor vibration signals, achieving real-time capture and anomaly diagnosis of idler vibration by laying detection optical fibers along the frame, which can accurately identify early faults. However, this method requires dense fiber optic deployment over long conveyor belt corridors, leading to a sharp increase in cost, and the single-mode signal is easily affected by environmental interference such as conveyor belt vibration and personnel movement. Therefore, given the reality of long-distance belt conveyor monitoring and severe wireless signal attenuation, there is an urgent need to develop a multi-mode idler monitoring device with anti-vibration coupling capabilities to ensure stable idler operation and improve the transport efficiency of belt conveyors. Summary of the Invention

[0004] To address the problems in the prior art, this invention provides an intelligent posture optimization pad system and its intervention method based on multimodal sensing and behavioral feedback.

[0005] In a first aspect, the present invention provides an intelligent posture optimization cushion system: The system comprises a core perception and processing unit, an interaction and execution unit, and an enterprise-level management and ecosystem interface unit. The specific structure of each unit is as follows: Core sensing and processing unit: Multimodal pressure sensor array module: Embedded inside the seat cushion, it consists of multiple high-precision piezoresistive sensors arranged in a matrix to collect pressure distribution image data of the user's buttocks and legs in real time and continuously.

[0006] Edge computing module: Integrated into the microcontroller inside the seat cushion, it is directly connected to the sensor array and has a built-in lightweight artificial intelligence inference model. It is used to process pressure data in real time on the device, perform posture feature extraction and poor posture pattern clustering and classification, and dynamically generate personalized intervention strategies based on the COM-B model.

[0007] Interaction and Execution Unit: Multimodal feedback execution module: Based on the intervention commands issued by the edge computing module, it provides users with real-time, non-invasive posture adjustment prompts in at least one way: a distributed vibration motor array that can simulate directional tactile cues; an RGB LED indicator group that conveys information through color and flashing patterns; and a wireless audio prompt unit that pairs with the user's headphones to provide voice or prompt sounds.

[0008] User-side application module: Installed on the user's smart terminal, it includes three types of functional modules: a gamified incentive module, designed based on behavioral psychology, which includes health goal setting, task challenges, points rewards, a virtual achievement system, and a social leaderboard; a team health visualization module, which, while protecting personal privacy, displays the overall sitting posture health data trends, compliance rates, and team challenge progress of the team (such as a department), creating a positive and healthy social atmosphere; and a personalized data dashboard, which displays the user's sitting posture habit analysis, improvement progress, health points, and personalized suggestions.

[0009] Enterprise-level management and ecosystem interface unit: Enterprise Health Management Interface Module: Deployed in the cloud or on the enterprise's local server, it mainly performs three functions: receiving aggregated and anonymized group health data uploaded from the user-end application module; data interface and integration with the enterprise's existing employee assistance program system, human resource management system, or enterprise health platform; and providing enterprise health managers with an organizational-level health data dashboard, including tools for group health risk analysis, intervention effect evaluation, and productivity loss cost-benefit analysis.

[0010] Data privacy and security module: Ensures that all sensitive raw stress data is processed on the device side, and only transmits fully anonymized analysis results (such as aggregated indicators such as the incidence of poor posture and average improvement time) to the cloud and enterprise side, and adopts encrypted communication protocols to fundamentally avoid the risk of data privacy leakage.

[0011] The training method for the lightweight AI inference model is as follows: a large amount of pressure distribution data covering various standard sitting postures (such as standard sitting posture, leaning forward, leaning back, leaning left, leaning right, crossing legs, etc.) and poor sitting postures of users of different body types and genders is collected, labeled, and then a neural network model with a small number of parameters and high computational efficiency is trained using transfer learning or knowledge distillation techniques to adapt to the resource limitations of the edge computing module.

[0012] Secondly, the present invention provides an intelligent sitting posture intervention method based on the above-mentioned system. The method includes the following steps: S1, Data Acquisition and Preprocessing: Raw data of pressure distribution under user sitting posture is acquired in real time through a multimodal pressure sensor array, and filtered and normalized preprocessing is performed to eliminate data noise and improve data validity. S2, Posture Feature Extraction and Pattern Recognition: In the edge computing module, a pre-trained lightweight convolutional neural network model is used to extract features from the preprocessed pressure distribution image; combined with time series analysis methods, the features of continuous frames are dynamically clustered to identify specific poor sitting posture patterns and durations, and severity is graded. S3, Personalized Intervention Strategy Generation and Execution: If a user's posture is identified as good, it is recorded and positive points are awarded to reinforce positive behavior. If poor posture is identified, the trigger threshold, feedback method, and feedback intensity of this intervention will be dynamically adjusted based on the user's historical behavioral data (such as response speed to reminders and improvement effects), current work status (such as calendar busy status accessed via API), and the evaluation results of the three dimensions (ability, opportunity, and motivation) of the COM-B model. For example, a gentler, guided feedback will be used for new users or users with low motivation; for users who have developed certain habits but occasionally slack off, the timeliness of reminders will be appropriately increased.

[0013] Based on the generated personalized strategy, the corresponding feedback execution module is triggered (such as a slight vibration in a specific area) to guide the user to adjust their sitting posture.

[0014] S4. Behavioral closed-loop incentives and data updates: Record users' responses to interventions (such as the time required for adjustment and whether the posture meets the standard after adjustment), update users' personal behavior profiles in real time, and provide data support for subsequent intervention strategy optimization; at the same time, provide immediate incentives through the gamification module (such as earning points for completing a correct adjustment) and social module (such as achieving the team's daily goal) of the user application to further strengthen users' positive posture behavior.

[0015] S5. Organizational Health Data Aggregation and Decision Support: Regularly upload anonymized user behavior data (such as total daily time spent in good sitting posture and number of times poor posture was corrected) to the enterprise health management interface module; this module performs multi-dimensional analysis of health trends at the enterprise as a whole or departmental level, generates visual reports, and provides data-driven decision-making basis for enterprises to formulate health welfare policies and carry out targeted health activities.

[0016] Furthermore, the specific implementation method of dynamically adjusting the intervention strategy based on the COM-B model includes: Capability Dimension: When the system detects that a user is repeatedly unable to adjust to the target posture correctly, it pushes short, targeted instructional videos or graphic guides through the APP to improve the user's posture adjustment skills, providing support for the user to improve their posture from a capability perspective.

[0017] Opportunity Dimension: Based on the user's calendar information, automatically switch to "Do Not Disturb" mode during important meetings or deep work sessions, recording without intervention to avoid disrupting the user's work; strengthen guidance during work breaks to enhance the intervention effect. Simultaneously, create opportunities for collective posture improvement through team challenges, fostering a healthy team atmosphere.

[0018] Motivation dimension: Combining points rewards, virtual badges, social recognition (team contribution value), and visualization of long-term health benefits (such as estimated reduction in pain days) to continuously stimulate and maintain users' motivation to improve their posture, and promote users to develop good posture habits from an intrinsic level.

[0019] Beneficial effects Compared with the prior art, the present invention has the following significant advantages: 1. Precise Identification: By adopting a high-density pressure sensor array and a lightweight AI model, it breaks through the limitation of existing products that only judge "prolonged sitting". It can accurately identify various specific poor sitting posture patterns such as forward tilting of the spine, scoliosis, and crossing legs, and classify the severity of poor sitting posture, providing a precise basis for personalized intervention.

[0020] 2. Personalized and Intelligent Intervention: Based on the COM-B behavior change theory, the intervention strategy's trigger threshold, feedback method, and intensity are dynamically adjusted by combining users' historical behavior data, real-time work status, and individual differences. This makes the reminders more tailored to users' needs, reduces user interference and "reminder fatigue," and improves the effectiveness of the intervention.

[0021] 3. Sustainable Incentives: Deeply integrate gamified incentives and social support mechanisms. Through points, badges, social leaderboards, team challenges, and other methods, it promotes users' long-term adherence to healthy sitting posture behavior from both intrinsic motivation and external environment, forming a closed loop of behavior change.

[0022] 4. Ecosystem Integration: It is the first to deeply integrate personal health intervention devices with corporate health management systems, making the product a digital interface for organizational health governance. This achieves a paradigm shift from "individual hardware" to "organizational solutions." It not only provides posture intervention for individuals but also provides data support for companies to formulate health welfare policies and reduce employee health risks, demonstrating significant social and commercial value.

[0023] 5. Privacy and security: Sensitive raw stress data is processed locally through edge computing, and the de-identified aggregated analysis results are only transmitted to the cloud and enterprise. Combined with encrypted communication protocols, the risk of health data privacy leakage is fundamentally solved, which complies with increasingly stringent data security regulations at home and abroad. Attached Figure Description

[0024] Figure 1 This is a schematic diagram of the overall architecture of the intelligent posture optimization cushion system of the present invention. Figure 2 This is a flowchart illustrating the closed-loop workflow of posture recognition and intervention in this invention.

[0025] Figure 3 This is a schematic diagram illustrating the logic for generating personalized intervention strategies based on the COM-B model of this invention.

[0026] Figure 4 This is a schematic diagram illustrating the flow and application of health data at the individual-team-organization level according to the present invention. Detailed Implementation

[0027] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0028] Example 1: System Hardware and Core Algorithm Implementation The core hardware of this system is a smart seat cushion. The seat cushion is embedded with a multimodal pressure sensor array consisting of 16×16 pressure sensor nodes, covering the main weight-bearing areas of the user's buttocks and legs. The main control chip uses a low-power microprocessor with a neural network acceleration unit to form an edge computing module, which is adapted to the hardware resource limitations of the seat cushion.

[0029] After the system is powered on, the pressure sensor array collects raw pressure distribution data from the user in real time at a frequency of 10Hz. The posture recognition algorithm is pre-installed in the edge computing module. It first calibrates and denoises the raw pressure data to generate a grayscale pressure distribution map, which is then input into an optimized MobileNetV2 lightweight convolutional neural network. This network has been trained on a database containing more than 5,000 hours of labeled data and can output the probability of the current sitting posture belonging to seven categories, including "standard", "forward tilt > 15 degrees", "left tilt", and "cross-legged".

[0030] Meanwhile, the algorithm calculates the trajectory of the pressure center point. If it shakes violently in a short period of time, it determines that the user is adjusting their posture on their own, and the system will temporarily suspend intervention to avoid ineffective reminders and improve the user experience.

[0031] Example 2: Implementation of Personalized Intervention Process When user Xiaoming first uses this system, the system initially uses the default medium sensitivity threshold. When the edge computing module detects that Xiaoming has been sitting in a "forward leaning" posture for more than 3 minutes, the intervention mechanism is triggered. Considering Xiaoming's new user status, the system selects a combination of "gentle voice prompts from the APP + slight vibrations at the front of the seat cushion" as the initial reminder.

[0032] Upon hearing the sound, Xiaoming adjusted his posture. The system detected that the pressure distribution had returned to the standard mode and immediately awarded him "+10 health points" as an instant reward through the APP, recording this positive behavior.

[0033] A week later, by studying Xiaoming's historical behavioral data, the system discovered that he was more sensitive to vibration feedback and more prone to leaning forward during afternoon fatigue periods. Therefore, during the afternoon, the system dynamically adjusted the tolerance threshold for leaning forward from 3 minutes to 2 minutes, primarily using vibration feedback for intervention. Simultaneously, the system found that Xiaoming had a low click-through rate on the knowledge tag "how to properly adjust seat height to reduce leaning forward," indicating a deficiency in his posture adjustment ability. Therefore, the system proactively pushed a 30-second educational video to him during his rest time via the app to improve his posture adjustment skills.

[0034] Example 3: Enterprise Application Implementation A technology company purchased 100 sets of this system and deployed them in the R&D department. All user data was anonymized and uploaded to the enterprise health management interface module after obtaining the user's consent.

[0035] Enterprise health managers saw through the organizational health data dashboard that the department's "total time spent in poor posture" decreased by 35% quarter-on-quarter in the second quarter, while the "proactive reporting rate of lower back discomfort" by employees decreased by 20%. The data also showed that the teams that participated in the "departmental health challenge" had significantly better posture improvement data than the teams that did not participate.

[0036] Based on the above data, the corporate health administrator officially incorporated this system into the annual employee health benefits package and plans to give extra health allowances to teams that have made significant improvements in sitting posture. This has created a virtuous cycle of "individual behavior improvement - team atmosphere creation - organizational policy support", achieving the coordinated development of individual health and corporate health management.

[0037] Example 4: Privacy Protection Implementation Throughout the entire data flow process, the user's original pressure distribution image never leaves the edge computing module of the smart cushion, and all sensitive data is processed on the device side. The edge computing module only synchronizes highly abstract data such as "event results" (e.g., 14:30-14:33, identified as leaning forward; 14:34, user successfully adjusted) to the user's mobile APP and the enterprise cloud through an encrypted communication channel.

[0038] This data processing method ensures that enterprises cannot obtain any real-time images of employees' specific sitting postures, nor can they infer users' personal identities and specific body postures through transmitted data, thus completely eliminating the risk of health data privacy leaks.

[0039] In addition, throughout this document, relational terms such as "first" and "second" are used merely to distinguish one entity or operation from another, without necessarily requiring or implying any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

Claims

1. An intelligent posture optimization cushion system, characterized in that, include: The core sensing and processing unit is used to collect user sitting posture pressure data and complete sitting posture recognition, classification and personalized intervention strategy generation locally. The interaction and execution unit communicates with the core perception and processing unit to provide users with feedback on posture adjustment and to implement user-side behavioral incentives and data visualization. The enterprise-level management and ecosystem interface unit communicates with the interaction and execution unit to aggregate, de-identify, and support enterprise-level health decision-making for group health data, while ensuring data privacy and security.

2. The intelligent posture optimization cushion system according to claim 1, characterized in that, The core sensing and processing unit includes: a multimodal pressure sensing array module, embedded inside the seat cushion, composed of high-precision piezoresistive sensors arranged in a matrix, used to collect pressure distribution image data of the user's buttocks and legs in contact areas in real time and continuously; and an edge computing module, integrated into a microcontroller inside the seat cushion and directly connected to the sensing array module, with a built-in lightweight artificial intelligence inference model, used to process pressure data in real time on the device, perform posture feature extraction, clustering and grading of poor posture patterns, and dynamically generate personalized intervention strategies based on the COM-B model.

3. The intelligent posture optimization cushion system according to claim 2, characterized in that, The lightweight AI inference model is trained through transfer learning or knowledge distillation techniques, adapts to the resource limitations of edge computing modules, and can identify various standard and poor sitting posture patterns.

4. The intelligent posture optimization cushion system according to claim 1, characterized in that, The interaction and execution unit includes: a multimodal feedback execution module, used to provide users with real-time, non-invasive posture adjustment prompts through at least one of the following methods: a distributed vibration motor array, an RGB LED indicator group, and a wireless audio prompt unit, based on the intervention instructions of the core perception and processing unit; and a user-end application module, installed on the user's smart terminal, including a gamified incentive module, a team health visualization module, and a personalized data dashboard, to realize the incentive of user posture behavior, team health data sharing, and personal posture data display.

5. The intelligent posture optimization cushion system according to claim 4, characterized in that, The gamified incentive module includes health goal setting, task challenges, points rewards, a virtual achievement system, and a social leaderboard; the team health visualization module displays the team's sitting posture health data trends and challenge progress while protecting personal privacy; and the personalized data dashboard displays user sitting posture habit analysis, improvement progress, and personalized suggestions.

6. The intelligent posture optimization cushion system according to claim 1, characterized in that, The enterprise-level management and ecosystem interface unit includes: an enterprise health management interface module, deployed in the cloud or on the enterprise's local server, used to receive anonymized group health data, integrate with the enterprise's existing management system, and provide enterprise managers with an organizational-level health data dashboard and decision analysis tools; and a data privacy and security module, used to restrict sensitive raw stress data to device-side processing, transmit only anonymized aggregated analysis results to the cloud and enterprise, and use encrypted communication protocols to ensure data transmission security.

7. The intelligent posture optimization cushion system according to claim 6, characterized in that, The desensitized aggregated analysis results include at least one of the following: incidence of poor sitting posture, average improvement time, total daily duration of good sitting posture, and number of times poor posture was corrected; the organization-level health data dashboard includes functions for population health risk analysis, intervention effect evaluation, and productivity loss cost-benefit analysis.

8. A smart posture intervention method based on the smart posture optimization cushion system according to any one of claims 1-7, characterized in that, Includes the following steps: S1. Data Acquisition and Preprocessing: Raw data on user sitting pressure distribution is acquired through a multimodal pressure sensor array module and preprocessed by filtering and normalization. S2. Posture Feature Extraction and Pattern Recognition: The edge computing module extracts posture features through a lightweight artificial intelligence inference model, and combines time series analysis to identify poor posture patterns, duration, and severity classification; S3. Personalized Intervention Strategy Generation and Execution: Based on the posture recognition results, positive incentives are given or personalized intervention strategies are generated based on the COM-B model, user historical behavior data, and current work status, and adjustment prompts are output through the multimodal feedback execution module; S4. Behavioral closed-loop incentives and data updates: Record user intervention response results, update personal behavior profiles, and provide real-time incentives through the user-end application module; S5. Organizational Health Data Aggregation and Decision Support: Upload anonymized user behavior data to the enterprise health management interface module, analyze it to generate visual reports, and provide data support for enterprise health management.

9. The intelligent sitting posture intervention method according to claim 8, characterized in that, In step S3, the method of generating personalized intervention strategies based on the COM-B model is as follows: Ability dimension: push instructional videos or graphic guides to users who have difficulty adjusting their sitting posture; Opportunity dimension: switch intervention modes according to the user's work status and create collective improvement opportunities through team challenges; Motivation dimension: stimulate user motivation by combining points rewards, virtual badges, social recognition and visualization of long-term health benefits.

10. The intelligent sitting posture intervention method according to claim 8, characterized in that, In step S2, if a violent shaking of the pressure center point trajectory is detected in a short period of time, it is determined that the user is in a posture adjustment state, and the system temporarily suspends the intervention operation. In step S3, gentle guiding feedback is used for new users or users with low motivation, and the timeliness of reminders is increased for users who have developed a sitting posture habit but occasionally relax.

Images