Seat feedback adjusting system supporting human body posture collection

Abstract

The invention relates to the technical field of seat adjustment, and discloses a seat feedback adjustment system supporting human body posture collection. A posture acquisition module of the system acquires pressure distribution data and timestamp information of a plurality of body parts of a user in real time; the partition processing module partitions the pressure data according to a preset rule to generate posture partition data; the posture calibration module extracts features and generates initial posture calibration, and marks a to-be-verified posture partition; the association verification module extracts feature vectors of the partitions to be verified, compares the feature vectors with a standard feature range to construct a feature sequence, verifies association conditions of feature pairs to be associated and counts the number of association pairs; the feedback adjustment module calculates a posture deviation degree evaluation value according to the number of the association pairs and the feature vectors which do not meet the association conditions, and generates a seat adjustment instruction. The system can accurately sense the real-time posture of the user, self-adaptive adjustment of the seat is achieved, and the use comfort is improved.

Description

Technical Field

[0001] This invention relates to the field of seat adjustment technology, specifically to a seat feedback adjustment system that supports human posture acquisition. Background Technology

[0002] In modern life and work settings, people spend increasingly more time seated. Whether it's an office chair, a lounge chair in the home, or a seat in public transportation, its fit to the human body directly impacts the user's comfort and health. Traditional chairs are typically designed based on average body measurements to determine their structure and dimensions, offering only simple height and backrest angle adjustments, or some chairs with manual adjustment of the seat cushion and backrest support. However, these adjustment methods have significant limitations.

[0003] Traditional chair adjustments lack precise perception of an individual's real-time posture. Different users have different heights, weights, and body proportions, and even the same user's sitting posture changes at different times, such as leaning forward while working or leaning back to rest. The pressure distribution on different parts of the body on the chair varies significantly in these different states. Traditional chairs cannot acquire this real-time pressure data, making it difficult for the adjusted position to consistently adapt to the user's actual posture. Users often need to make frequent manual adjustments, which is not only cumbersome but also makes it difficult to find the most suitable support position.

[0004] Most existing chairs with some sensing capabilities can only collect pressure information from a single body part, such as the pressure value at the front of the seat cushion or the middle of the backrest, failing to comprehensively acquire pressure distribution across multiple body areas. This single sensing method results in a one-sided judgment of posture, unable to accurately identify whether the user has poor posture, such as unsupported lower back, excessive pressure on the buttocks, or insufficient leg support. Even some products that can collect pressure data from multiple areas lack effective processing and analysis of this data, failing to transform scattered pressure information into an accurate judgment of posture, and thus unable to achieve targeted adjustments.

[0005] Current seat adjustment processes lack dynamic verification and correlation analysis of posture. A user's sitting posture is a dynamic process; changes in posture in one area of ​​the body can trigger a chain reaction in other areas. For example, leaning forward can alter the pressure distribution on the back, while the pressure of the legs on the seat cushion will also adjust accordingly. Traditional chairs cannot verify the correlation of these dynamic posture changes and can only adjust based on information from a single area. This can easily lead to an overall imbalance in posture after adjustment, reducing the comfort of the chair and potentially exacerbating strain on the cervical and lumbar spine with long-term use.

[0006] As people's demands for health and comfort continue to increase, the shortcomings of traditional seats in terms of posture perception and adaptive adjustment are becoming increasingly apparent, failing to meet users' needs for personalized and intelligent seats. Therefore, developing a seat system that can accurately collect human posture data, effectively analyze posture status, and achieve adaptive feedback adjustment has become an urgent problem to be solved in the current seat technology field. Summary of the Invention

[0007] The purpose of this invention is to provide a seat feedback adjustment system that supports human posture acquisition, so as to solve the problems mentioned in the background art.

[0008] To achieve the above objectives, the present invention provides a seat feedback adjustment system that supports human posture acquisition, the system comprising: The posture acquisition module is used to acquire pressure distribution data and corresponding timestamp information of multiple body parts of the user on the seat in real time; The partitioning module is used to partition the pressure distribution data acquired by the posture acquisition module according to the preset body region division rules, and generate posture partition data corresponding to each body region. The pose calibration module is used to extract features from the pose partition data generated by the partition processing module, and generate an initial pose calibration for each body region based on the extraction results. When a pose calibration to be verified is detected, the pose partition data of the corresponding body region is marked as the pose partition to be verified. The association verification module is used to extract the partition feature vectors of all pose partitions to be verified, compare each partition feature vector with the preset standard pose feature range, construct a first feature sequence and a second feature sequence based on the comparison results, serialize the feature vectors in the first feature sequence and the second feature sequence, and take two adjacent feature vectors as feature pairs to be associated, verify whether the feature pairs to be associated meet the association conditions based on the pose association model, and count the number of association pairs that meet the association conditions. The feedback adjustment module is used to calculate the posture deviation evaluation value based on the number of association pairs output by the association verification module and the feature vectors that do not meet the association conditions, and to generate a seat adjustment command based on the posture deviation evaluation value.

[0009] Preferably, the body part pressure distribution data acquired by the posture acquisition module includes pressure value, pressure distribution coordinates, and pressure duration; The preset body region division rules of the partitioning module include the coordinate range definition of the head support area, back support area, waist support area and leg support area.

[0010] Preferably, the specific process of feature extraction by the posture calibration module is as follows: calculate the mean pressure, pressure change gradient and pressure distribution dispersion of each posture partition data, and combine them to generate the partition feature vector; The specific process for generating the initial pose calibration is as follows: compare each partition feature vector with a preset standard feature vector threshold. If all dimensions of the partition feature vector are within the standard feature vector threshold range, the generated pose conforms to the calibration. If all dimensions exceed the threshold, the generated pose deviates from the calibration. If some dimensions exceed the threshold, a pose calibration to be verified is generated.

[0011] Preferably, the specific process by which the association verification module constructs the first feature sequence and the second feature sequence is as follows: When the number of deviation dimensions of the partition feature vector is less than a preset threshold, it is assigned to the first feature sequence; When the number of deviation dimensions of the partition feature vector is higher than or equal to a preset threshold, it is assigned to the second feature sequence; The specific process of the serialization is as follows: sort the partition feature vectors in each feature sequence according to the timestamp order.

[0012] Preferably, the specific process of the association verification module based on the pose association model verification is as follows: The partition feature vectors and correlation verification results of historical pose partition data are used as the training sample set. The pose association model is trained using a temporal feature matching algorithm; The sorted features to be associated are input into the pose association model, and the association matching degree is output. When the correlation matching degree exceeds the preset matching threshold, the correlation condition is determined to be met.

[0013] Preferably, the specific process by which the feedback adjustment module calculates the posture deviation evaluation value is as follows: Count the number of associated pairs and the number of unassociated feature vectors in the first feature sequence that satisfy the association condition; Count the number of associated pairs and the number of unassociated feature vectors that satisfy the association condition in the second feature sequence; The number of associated pairs and the number of unassociated feature vectors are weighted and fused according to preset weight coefficients.

[0014] Preferably, the specific process by which the feedback adjustment module generates seat adjustment commands is as follows: The posture deviation assessment value is matched with multiple preset adjustment level thresholds; Based on the matching results, commands to adjust the seat height, backrest angle, or lumbar support intensity at the corresponding level are triggered.

[0015] Preferably, the system further includes: The resource scheduling module is used to monitor the real-time computing load of the posture acquisition module, partition processing module, posture calibration module, correlation verification module and feedback adjustment module. The dynamic allocation module is used to dynamically allocate processing tasks to edge computing nodes based on the computing load data monitored by the resource scheduling module through an index mapping mechanism.

[0016] Preferably, the specific process of the dynamic allocation module is as follows: Establish task processing capability indexes for each processing module and resource status indexes for edge computing nodes; A bidirectional index traversal algorithm is used to match task processing requirements with resource availability. Based on the matching results, the pose partitioning data processing task is assigned to the target edge computing node in real time.

[0017] Preferably, the system further includes a load balancing module, which is used to dynamically reclaim or allocate computing resources of each edge computing node according to the task allocation records of the dynamic allocation module through an elastic resource pool adjustment algorithm.

[0018] Compared with the prior art, the beneficial effects of the present invention are: In terms of the comprehensiveness and real-time nature of posture acquisition, the posture acquisition module in the system can acquire pressure distribution data and corresponding timestamp information for multiple body parts of the user on the seat in real time. Compared with traditional seats that can only achieve simple manual adjustment or single-part pressure sensing, this module can comprehensively capture the pressure of each body part in contact with the seat during changes in the user's sitting posture. Whether it is the pressure on the buttocks and legs in the seat cushion area, or the pressure on the back and waist in the backrest area, it can all be recorded in real time. The addition of timestamp information can also accurately reflect the posture change patterns at different time points, providing a complete data foundation for subsequent analysis of the user's dynamic changes in sitting posture, and avoiding posture judgment errors caused by incomplete or delayed data acquisition. Regarding the accuracy of pressure data processing, the partitioning module divides the pressure distribution data into zones according to preset body region division rules, generating posture partition data corresponding to each body region. This partitioning method systematically categorizes the originally scattered pressure data, making the pressure information for each body region clearer and more precise, avoiding the analysis difficulties caused by mixed pressure data from multiple regions. For example, the seat cushion can be divided into a hip support zone and a leg support zone, and the backrest can be divided into a shoulder support zone and a lumbar support zone, etc. The pressure data of each zone is grouped independently, providing precise data units for subsequent posture analysis and adjustment for different body parts, ensuring that the analysis of sitting posture can be refined to specific body regions, rather than remaining at the level of a vague overall judgment. The posture calibration module extracts features from the data of each posture region and generates an initial posture calibration. It also marks the posture regions to be verified, effectively solving the problem that traditional chairs cannot accurately identify individual differences in sitting posture. Different users have different body characteristics and sitting habits. The initial posture calibration can establish a unique posture benchmark based on the user's real-time pressure data, rather than relying on a fixed universal standard. When a posture calibration to be verified is identified, the corresponding area of ​​data is marked as the region to be verified, avoiding misjudgments due to accidental posture fluctuations and ensuring a more rigorous judgment of sitting posture. For example, when a user experiences abnormal pressure data in a certain area due to a brief adjustment of their body, the module will not directly classify it as poor sitting posture, but will further confirm it through subsequent correlation verification, improving the accuracy of posture judgment. The correlation verification module extracts the feature vectors of the posture regions to be verified, compares them with the preset standard posture feature range to construct a feature sequence, and performs serialization processing on the feature vectors to verify whether the feature pairs to be correlated meet the correlation conditions, thus realizing dynamic correlation analysis of sitting posture. This analysis method fully considers the holistic nature of human sitting posture; changes in posture in one body area will be correlated with other areas. For example, abnormal lumbar posture may be accompanied by changes in the pressure distribution of the back and buttocks. By verifying the correlation conditions of the feature pairs to be correlated, the mutual influence between postures of different body areas can be accurately identified, and the number of correlation pairs that meet the correlation conditions can be counted. This allows for a comprehensive judgment on whether the current sitting posture is in an overall balanced and comfortable state, avoiding the overall posture imbalance problem caused by traditional chairs that adjust based on information from only a single area. The feedback adjustment module calculates the posture deviation assessment value based on the number of correlation pairs and feature vectors that do not meet the correlation conditions, and generates seat adjustment commands, making seat adjustment more targeted and scientific. The posture deviation assessment value can quantify the difference between the current sitting posture and the standard comfortable sitting posture, and clearly define the direction and degree of adjustment required, rather than the blind adjustment of traditional seats. For example, when the assessment value shows a high posture deviation in the lumbar region, and correlation verification finds that the feature vectors of the lumbar and back do not meet the correlation conditions, the feedback adjustment module can accurately generate commands to adjust the position or firmness of the backrest lumbar support. This allows the seat to dynamically adjust according to the user's real-time posture, continuously maintaining a good fit with the user's body, effectively improving the comfort of the seat, reducing the risk of physical strain caused by long-term poor posture, and meeting users' needs for personalized and intelligent seats. Attached Figure Description

[0019] Figure 1 This is a timing diagram of the seat feedback adjustment system supporting human posture acquisition as described in this invention; Figure 2 A flowchart for acquiring body part data and dividing the region; Figure 3A flowchart for constructing and serializing a feature sequence; Figure 4 A flowchart for calculating the posture deviation assessment value. Detailed Implementation

[0020] 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.

[0021] Please see Figure 1 This invention provides a seat feedback adjustment system that supports human posture acquisition, the system comprising: The system works collaboratively. The posture acquisition module is responsible for acquiring real-time pressure distribution data and corresponding timestamps for multiple body parts of the user on the seat. This module captures data through a built-in pressure sensor array to ensure continuous monitoring of user posture changes. The partitioning module receives the pressure distribution data and partitions it according to preset body region division rules, generating posture partition data for each body region. Body regions are typically divided into head support, back support, lumbar support, and leg support areas, each with clearly defined coordinate ranges to ensure targeted data processing. The posture calibration module extracts features from the posture partition data output by the partitioning module, calculates the mean pressure, pressure change gradient, and pressure distribution dispersion for each partition, and generates initial posture calibration based on these features.

[0022] If the feature extraction results show that some dimensions exceed the preset standard, the posture partition data of that body region is marked as a posture partition to be verified. The association verification module processes all posture partitions to be verified, extracts their partition feature vectors, and compares these vectors with the preset standard posture feature range. Based on the comparison results, the feature vectors are classified into a first feature sequence and a second feature sequence, serialized, and sorted by timestamp to form feature pairs to be associated. The posture association model is used to verify whether these feature pairs meet the association conditions and counts the number of associated pairs that meet the conditions. The feedback adjustment module calculates the posture deviation evaluation value based on the output of the association verification module. This value is obtained by weighted fusion of the number of associated pairs and the number of unassociated feature vectors, and generates seat adjustment instructions based on the evaluation value, such as adjusting the seat cushion height, backrest angle, or lumbar support strength. The entire system achieves real-time posture monitoring and feedback through modular design, improving the user experience.

[0023] Example 1: See Figure 2The posture acquisition module uses an array of pressure sensors integrated into the seat surface to acquire real-time pressure distribution data from multiple body parts of the user. These sensors are arranged in a grid to detect the pressure exerted on the seat by different parts of the user's body when seated. Pressure value refers to the specific numerical value of pressure measured by the sensors, that is, the magnitude of the pressure, usually measured in Newtons or Pascals. For example, when a user sits down, the sensors record the pressure value at each point, which helps to understand which areas experience greater or less pressure. Pressure distribution coordinates refer to the specific position coordinates of each pressure sensor on the seat. These coordinates are based on a preset coordinate system, such as establishing a coordinate system with the lower left corner of the seat as the origin. Each sensor has a unique coordinate identifier, thus accurately mapping the source location of the pressure data. Pressure duration records the length of time the pressure value remains constant, from the start of pressure detection to the end of the time interval. This time information is used to analyze the stability or frequency of posture changes. This data is collected along with timestamp information, which indicates the specific time point of data acquisition, ensuring the temporal order of the data.

[0024] After receiving this pressure distribution data, the partitioning module processes it according to preset body region division rules. These rules, designed based on ergonomic principles, divide the seat surface into multiple regions, each corresponding to a specific part of the body. The head support zone typically covers the top of the seat back, though the exact coordinates may vary depending on the seat size. The back support zone is located below the head support zone, covering the entire back contact surface. The lumbar support zone focuses on the lumbar spine area and is usually a smaller area. The leg support zone includes the seat cushion, covering the thighs and buttocks. Based on these coordinate ranges, the partitioning module divides the raw pressure data into different data blocks, each corresponding to a body region, and adds timestamp information to form posture partitioning data. This process ensures targeted and efficient data processing, allowing subsequent modules to focus on analyzing specific regions.

[0025] The posture calibration module extracts features from the posture partitioning data generated by the partitioning module. The aim is to extract meaningful indicators from the raw pressure data to assess posture status. When calculating the pressure mean, the module takes the arithmetic mean of all pressure values ​​within a body region. The pressure mean represents the average pressure intensity of that region, i.e., the average level of pressure magnitude. The pressure change gradient calculates the rate of change of pressure values ​​per unit time, approximating the derivative using the finite difference method. This gradient value helps identify dynamic changes in pressure, such as rapid changes in pressure when the user moves. The pressure distribution dispersion is obtained by calculating the standard deviation or variance of the pressure values, representing the uniformity of the pressure distribution. These features are combined into a partition feature vector, which is typically a three-dimensional or multi-dimensional array, with each dimension corresponding to a feature value.

[0026] The process of generating the initial pose calibration involves comparing the partition feature vector with a preset standard feature vector threshold. The standard feature vector threshold is set based on an ideal pose model, with each feature dimension having a minimum and maximum threshold range. During the comparison, the module checks whether the value of each dimension of the partition feature vector is within the corresponding threshold range. If all dimension values ​​are within the threshold range, the module generates a pose that conforms to the calibration, indicating that the pose of that body region is normal. If all dimension values ​​exceed the threshold range, the module generates a pose that deviates from the calibration, indicating an abnormal pose. If some dimension values ​​exceed the threshold range, the module generates a pose calibration to be verified, indicating that the pose is uncertain and requires further verification. After the pose calibration to be verified is triggered, the pose partition data of the corresponding body region is marked as the pose partition to be verified for subsequent module processing. This calibration mechanism allows the system to efficiently handle uncertainties.

[0027] Data flow and coordination between modules are crucial. The posture acquisition module continuously outputs a data stream, and the partition processing module processes this data in real time to ensure low latency. The feature extraction algorithm of the posture calibration module needs to be optimized for computational efficiency to avoid system bottlenecks. The calculation of the pressure mean can use the sliding window averaging method to reduce the computational burden. The calculation of the pressure change gradient may involve numerical differentiation methods, such as the central difference method, to improve accuracy. The calculation of the pressure distribution dispersion relies on statistical functions, such as the standard deviation formula. The standard feature vector threshold is set based on a large amount of experimental data or industry standards, and may be derived through machine learning training. During calibration generation, the module may use conditional statements or lookup tables to achieve fast comparison. The labeling of the posture partitions to be verified is accomplished by setting flag bits or creating independent data structures, facilitating access by the associated verification module. The entire implementation emphasizes practicality and scalability, allowing integration into various seating systems.

[0028] Example 2: See Figure 3When processing the posture partitions to be verified, the correlation verification module extracts the partition feature vectors of all posture partitions. These vectors contain feature values ​​such as mean pressure, pressure change gradient, and pressure distribution dispersion. Each partition feature vector needs to be compared with a preset standard posture feature range. The standard posture feature range is a reference value range derived from a large amount of normal posture data. Each feature dimension has a corresponding lower and upper limit value. For example, the standard range for mean pressure might be 100 to 200 Pascals, and the standard range for pressure change gradient might be -5 to 5 Pascals / second. During the comparison process, the module calculates the degree of deviation of each dimension value in each partition feature vector from the standard range. The degree of deviation is obtained by simple arithmetic difference. For example, if a feature value is 220 Pascals, and the standard upper limit is 200 Pascals, then the deviation value is 20 Pascals. The number of deviation dimensions refers to the number of dimensions in the partition feature vector that exceed the standard range. This is a counting indicator. For example, if a vector has three dimensions, and two of them exceed the standard range, then the number of deviation dimensions is 2.

[0029] The preset threshold is an integer parameter, typically set empirically to half the total number of dimensions or a specific value, used to distinguish different degrees of deviation. When the number of deviation dimensions in a partition feature vector is less than the preset threshold, it is assigned to the first feature column; these vectors represent slight deviations, possibly caused by temporary movements or measurement noise. When the number of deviation dimensions is greater than or equal to the preset threshold, it is assigned to the second feature column; these vectors represent significant deviations, potentially indicating true pose problems. The serialization process involves sorting the partition feature vectors in each feature column according to their timestamps. The timestamp information is obtained from the pose acquisition module to ensure temporal consistency. After sorting, the first and second feature columns form ordered sequences, where each element is a partition feature vector. After serialization, adjacent feature vectors are used as feature pairs to be associated; for example, the i-th vector and the (i+1)-th vector in the sequence form a pair. These feature pairs are used for subsequent association verification.

[0030] The pose association model is the core component of the association verification module, used to verify whether feature pairs to be associated meet the association conditions. The model's training process is based on historical pose partitioning data, including partition feature vectors and corresponding association verification results. These results are derived from manual annotation or automated judgment, with annotations indicating which feature pairs are truly associated. The training sample set needs to cover multiple pose scenarios to ensure the model's generalization ability. A temporal feature matching algorithm is used to train the pose association model. This algorithm analyzes the temporal correlation between feature vectors, for example, using dynamic time warping to align time series or recurrent neural networks to capture temporal dependencies. During training, the model learns to identify patterns between feature pairs, such as similar trends or stable relationships. After training, the model outputs an association matching degree, a numerical metric representing the association strength between two feature vectors. The sorted feature pairs to be associated are input into the pose association model, which calculates the association matching degree for each pair and outputs a scalar value. The preset matching threshold is an empirical value, set based on model performance, for example, by determining the optimal threshold point through ROC curve analysis. When the association matching degree exceeds the preset matching threshold, the association condition is considered met; otherwise, it is considered not met. The number of association pairs that meet the association criteria is counted for subsequent evaluation. The formula for calculating the association matching degree is: , in: This represents the degree of correlation matching and is a value between 0 and 1. Indicates the number of dimensions of the feature vector; and These represent the k-th dimension values ​​of the first and second feature vectors, respectively. This formula calculates the average similarity of the relative differences across each dimension of the two feature vectors; a higher value indicates a stronger correlation. This process improves verification accuracy and reduces subjective misjudgments through quantitative analysis.

[0031] The association verification module needs to efficiently process large amounts of data. When extracting partition feature vectors, the module may use batch processing to reduce computational overhead. When comparing standard pose feature ranges, vectorization operations are employed to improve speed, such as using parallel computing to compare multiple dimensions. Calculations for deviations from the specified dimensionality are implemented through conditional counting to quickly determine the sequence to which the model belongs. Serialization processing utilizes timestamp sorting algorithms, such as quicksort or radix sort, to ensure temporal correctness. The deployment of the pose association model may employ an embedded inference engine to optimize runtime efficiency. The entire process emphasizes real-time performance and is suitable for resource-constrained environments, such as embedded systems.

[0032] Example 3: See Figure 4When calculating the posture deviation assessment value, the feedback adjustment module needs to count the number of correlated pairs and the number of uncorrelated feature vectors in the first feature series. The number of correlated pairs refers to the total number of feature pairs confirmed by the correlation verification module to meet the conditions. These feature pairs represent posture data that are correlated in the time series, that is, there is a certain temporal correlation between these data points, indicating that the user's posture changes have a certain regularity or continuity. The number of uncorrelated feature vectors refers to the number of feature vectors in the first feature series that failed to form a correlated pair or failed to be correlated. These vectors may be due to isolated events or measurement noise, which cannot establish an effective temporal relationship with other data points, such as abnormal data points caused by sudden brief movements of the user or instantaneous interference of the sensor. Similarly, the module counts the number of correlated pairs and the number of uncorrelated feature vectors in the second feature series that meet the correlation conditions. The statistical method for the second feature series is similar, but because it contains feature vectors with a higher degree of deviation, these data usually represent more significant posture problems, and therefore will be assigned different importance weights in subsequent calculations. The statistical process employs a counting algorithm. It iterates through all feature pairs and feature vectors, accumulating counts based on the verification results. Specifically, this includes initializing a counter variable, iterating through each feature pair and checking its association status, updating the corresponding counter variable based on the check results, and outputting the statistical results. These statistical results provide the foundational data input for subsequent weighted fusion calculations, ensuring that the evaluation process is based on comprehensive data analysis.

[0033] The number of associated pairs and the number of unassociated feature vectors are weighted and fused according to preset weighting coefficients. These weighting coefficients are numerical parameters based on experimental data and analysis, used to adjust the contribution of different data components to the final evaluation value. These coefficients are optimized through extensive experimental data and can accurately reflect the actual influence of different data components. Typically, the first feature series has a lower weighting coefficient because the slight deviations it represents may have a smaller impact, such as temporary posture adjustments or minor changes in movement; the second feature series has a higher weighting coefficient because the significant deviations it represents require more attention, such as persistent poor posture or abnormal posture. The weighted fusion calculation uses a linear combination formula to combine data from different sources according to importance into a comprehensive evaluation value. This calculation process can comprehensively consider the overall influence of various factors. The calculation formula is: , in: This represents the final calculated posture deviation assessment value, which is a comprehensive quantitative indicator. and These represent the overall weight coefficients of the first and second characteristic sequences, respectively, and are used to adjust the overall contribution of different sequences; and These represent adjustment coefficients for unrelated feature vectors, used to finely adjust the degree of influence of unrelated data; These represent the number of associated pairs and the number of unassociated feature vectors in the first feature sequence, respectively. These represent the number of associated pairs and the number of unassociated feature vectors in the second feature sequence, respectively. The calculated pose deviation assessment value is a quantitative indicator; the higher the value, the more severe the pose deviation. This value can objectively reflect the overall degree of abnormality in the user's pose.

[0034] The process of generating seat adjustment commands based on posture deviation assessment values ​​involves matching the assessment values ​​with multiple preset adjustment level thresholds. These thresholds are predefined numerical ranges that divide consecutive assessment values ​​into different level intervals, such as low deviation, medium deviation, and high deviation. Each level interval corresponds to different adjustment strategies and parameter settings. The matching process is achieved by comparing the assessment values ​​with the threshold ranges. When an assessment value falls into a certain level interval, the corresponding adjustment command is triggered. This process is completed using an interval judgment algorithm. The seat height adjustment command changes the seat height by controlling the electric lifting mechanism, thereby adjusting the user's leg posture and improving lower limb blood circulation. The backrest angle adjustment command changes the backrest tilt angle by driving the backrest tilt motor, optimizing back support and reducing spinal pressure. The lumbar support strength adjustment command provides personalized lumbar support and enhances lumbar spine protection by adjusting the air pump pressure or the degree of mechanical protrusion. After these commands are generated, they are sent to the corresponding actuators via a control bus to achieve automatic seat adjustment. The entire adjustment process forms a closed-loop control, dynamically adjusting seat parameters based on real-time posture assessment results to provide users with continuous posture support and health protection. The system also includes a safety monitoring mechanism to ensure a smooth and reliable adjustment process, avoiding user discomfort caused by sudden changes in posture.

[0035] Example 4: The resource scheduling module continuously monitors the real-time computational load of the posture acquisition module, partition processing module, posture calibration module, correlation verification module, and feedback adjustment module. The computational load data includes performance indicators such as CPU utilization, memory usage, and processing latency for each module. These indicators are collected at a fixed sampling frequency, e.g., once per second, using the system's built-in performance monitoring tool. CPU utilization represents the percentage of processor resources used, indicating the processor's workload; a higher value indicates a heavier computational task. Memory usage reflects memory usage, referring to the amount of memory space currently occupied by the module, typically measured in megabytes. Processing latency refers to the time required for data processing, the time interval from data input to output; a shorter latency indicates higher processing efficiency. During monitoring, the resource scheduling module maintains a real-time data table recording the load status of each module. This table includes timestamps, module identifiers, and various indicator values. The timestamp records the specific time of data acquisition, the module identifier distinguishes different processing modules, and the indicator values ​​objectively reflect the operating status of each module. The module operation load monitoring index table shown below displays typical monitoring data, which comes from sampling records during actual system operation. See Table 1.

[0036] Table 1: Module Computational Load Monitoring Indicators

[0037] The dynamic allocation module dynamically distributes processing tasks to edge computing nodes based on real-time computing load data provided by the resource scheduling module, using an index mapping mechanism. This index mapping mechanism is a data management method that achieves task allocation by establishing two key indexes: a task processing capacity index for each processing module and a resource status index for each edge computing node. The task processing capacity index records the computing requirements of each module, including parameters such as the required number of CPU cores, memory size, and network bandwidth requirements. These parameters reflect the basic resource conditions required for module operation. The resource status index records the real-time available resources of each edge computing node, including the number of idle CPU cores, remaining memory capacity, and current network throughput. This information reflects the node's current processing capacity. Both indexes are stored using a key-value pair data structure for easy and fast querying and updating. The key is typically a unique identifier for the module or node, while the value stores the corresponding resource parameters.

[0038] A bidirectional index traversal algorithm matches task processing requirements with resource availability. This algorithm simultaneously traverses the task processing capacity index and the resource status index, calculating a matching score for each task-node pair. The matching score is calculated based on factors such as resource sufficiency and network latency, selecting the node with the highest score as the target allocation node. Based on the matching results, the posture partitioning data processing task is allocated to the target edge computing node in real time. The task allocation process includes packaging task data into packets, transmitting them over the network to the target node, receiving and processing the data, and returning the results. The entire allocation process emphasizes low latency and high reliability to ensure that the system's real-time requirements are met. Edge computing nodes are typically deployed close to the network edge of the seating device to reduce communication latency. After task execution, the resource status index is updated promptly to reflect the latest resource availability. This dynamic allocation mechanism enables the system to flexibly respond to load fluctuations and improve overall resource utilization.

[0039] The resource scheduling module employs multi-threading technology to collect load data from each module in parallel. Each monitoring indicator has a set safety threshold range; when an indicator exceeds the threshold, an early warning mechanism is triggered. The dynamic allocation module's index update frequency is consistent with the data collection frequency, ensuring that resource allocation decisions are based on the latest system status. Edge computing nodes are equipped with dedicated task processing engines that can quickly parse and process allocated posture partition data, and the processing results are returned to the main system through an encrypted channel. The entire resource scheduling and task allocation process forms a complete closed-loop control, which can automatically adjust the resource allocation strategy according to real-time load conditions, ensuring stable system operation under various operating conditions. The system also features a redundancy backup mechanism; when an edge computing node fails, tasks can be automatically redistributed to other available nodes, ensuring uninterrupted service. This design gives the system good scalability and fault tolerance, enabling it to adapt to application scenarios of different scales.

[0040] Example 5: The specific process of dynamically allocating modules includes establishing a task processing capacity index for each processing module and a resource status index for edge computing nodes. The task processing capacity index is a data structure that records the computing requirements of each module, including various resource parameters required for module operation, such as the number of processor cores, the amount of memory space required, and the required network connection speed. The resource status index is a data table that records the current available resources of each edge computing node, updating in real time how many idle processors, how much memory capacity, and the current network transmission capability of each node. Both indexes store data in key-value pairs, where the key is the module or node identifier and the value is the corresponding resource parameter value. A bidirectional index traversal algorithm is used to match task processing requirements with resource availability. This algorithm simultaneously checks the contents of both indexes and calculates a matching score between each task and each node. This score reflects whether the node has the resource conditions required to process the task. The matching score is calculated using the following formula: , in: This represents the match score; a higher score indicates a better match. , , These are the weighting coefficients for CPU, memory, and bandwidth, which are determined based on the importance of the task type. and These represent the number of available CPU cores for a node and the number of CPU cores required by a task, respectively. and These represent the available memory capacity of the node and the memory capacity required by the task, respectively. and These represent the available bandwidth of the node and the bandwidth required by the task, respectively. Based on the matching results, the pose partitioning data processing task is allocated to the target edge computing node in real time. This process includes packaging the data to be processed and sending it to the selected node. After receiving the data, the node performs the computation and processing and returns the processing result to the main system.

[0041] The load balancing module dynamically reclaims or allocates computing resources to each edge computing node based on the task allocation records from the dynamic allocation module, using an elastic resource pool adjustment algorithm. The task allocation records are log files generated by the dynamic allocation module during task allocation, detailing information such as the time of each allocation, which tasks were allocated, which node they were assigned to, and the size of the tasks. The elastic resource pool adjustment algorithm analyzes this historical data to predict future resource demand trends and dynamically adjusts the resource allocation of each node based on the prediction results. This algorithm may employ machine learning methods to improve prediction accuracy, such as using time series analysis algorithms to predict changes in resource demand. Specific operations for dynamically reclaiming or allocating computing resources include increasing or decreasing the number of processor cores allocated to a node, adjusting memory allocation, or changing network bandwidth configurations. The entire adjustment process is fully automated, requiring no manual intervention; the system automatically makes optimal resource allocation decisions based on real-time load conditions. The load assessment value is a comprehensive indicator of a node's current workload, considering factors such as CPU utilization, memory utilization, and network utilization. The system uses this assessment value to determine whether resource allocation adjustments are necessary.

[0042] The dynamic allocation module updates the contents of the two indexes at a fixed frequency, typically consistent with the resource monitoring frequency, ensuring that allocation decisions are based on the latest system status. The load balancing module's elastic resource pool adjustment algorithm runs periodically, analyzing task allocation records over a recent period and adjusting resource allocation strategies based on the analysis results. Edge computing nodes are equipped with resource monitoring agents, which collect node resource usage in real time and report it to the main system. When the system detects that a node is overloaded, it automatically transfers some tasks to less loaded nodes. Conversely, when a node is underloaded, the system may transfer tasks from other nodes to improve resource utilization. This dynamic adjustment mechanism ensures the system maintains a good load balance, avoiding imbalances where some nodes are overloaded while others are idle. The system also incorporates a fault-tolerance mechanism; when a node fails, its tasks are automatically redistributed to other healthy nodes to ensure uninterrupted service. The entire resource management process forms a complete closed-loop control system, automatically adjusting resource configuration based on real-time operating status to guarantee stable and efficient system operation under various conditions. This design gives the system excellent scalability and adaptability, allowing for flexible increases or decreases in the number of computing nodes as needed without affecting overall system performance. The resource monitoring agent is a software program running on each computing node that collects node operational data and sends it to the central management system. This data includes metrics such as processor usage, memory usage, and network traffic.

[0043] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply 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.

[0044] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A seat feedback adjustment system supporting human posture acquisition, characterized by, The method comprises the following steps: A posture acquisition module is used to acquire the pressure distribution data of multiple body parts of a user on a seat and corresponding timestamp information in real time; A partition processing module is used to partition the pressure distribution data acquired by the posture acquisition module according to a preset body region division rule to generate posture partition data corresponding to each body region; A posture calibration module is used to extract features from the posture partition data generated by the partition processing module and generate an initial posture calibration for each body region according to the extraction results, and when it is identified that there is a posture calibration to be verified, the posture partition data of the corresponding body region is marked as posture partition to be verified; An association verification module is used to extract the partition feature vectors of all posture partitions to be verified, compare each partition feature vector with a preset standard posture feature range, construct a first feature sequence and a second feature sequence according to the comparison results, perform sequence processing on the feature vectors in the first feature sequence and the second feature sequence, and take two adjacent feature vectors as a feature pair to be associated, verify whether the feature pair to be associated meets an association condition based on a posture association model, and count the number of associated pairs that meet the association condition; A feedback adjustment module is used to calculate a posture deviation degree evaluation value based on the number of associated pairs output by the association verification module and the feature vectors that do not meet the association condition, and generate a seat adjustment instruction based on the posture deviation degree evaluation value.

2. The seat feedback adjustment system for supporting human posture acquisition according to claim 1, wherein, The body part pressure distribution data acquired by the posture acquisition module includes pressure values, pressure distribution coordinates and pressure duration times; The body region division rule preset by the partition processing module includes the coordinate range definitions of the head support area, the back support area, the waist support area and the leg support area.

3. The seat feedback adjustment system for supporting posture acquisition of a human body according to claim 2, wherein The specific process of feature extraction by the posture calibration module is to calculate the pressure mean value, pressure change gradient and pressure distribution dispersion of each posture partition data, and combine to generate the partition feature vector; The specific process of generating the initial posture calibration is to compare each partition feature vector with a preset standard feature vector threshold value, generate a posture compliance calibration if all dimensions of the partition feature vector are within the standard feature vector threshold value range, generate a posture deviation calibration if all dimensions exceed the threshold value, and generate a posture calibration to be verified if some dimensions exceed the threshold value.

4. The seat feedback adjustment system for supporting posture acquisition of a human body according to claim 3, wherein The specific process of the association verification module for constructing the first feature sequence and the second feature sequence is as follows: When the number of deviated dimensions of the partition feature vector is lower than a preset threshold value, it is classified into the first feature sequence; When the number of deviated dimensions of the partition feature vector is higher than or equal to the preset threshold value, it is classified into the second feature sequence; The specific process of the sequence processing is to sort the partition feature vectors in each feature sequence in the order of timestamps.

5. The seat feedback adjustment system for supporting posture acquisition of a human body according to claim 4, wherein The specific process of the association verification module based on the posture association model is as follows: The partition feature vectors and association verification results of historical posture partition data are acquired as a training sample set; A time sequence feature matching algorithm is used to train the posture association model; The sorted feature pairs to be associated are input into the posture association model, and an association matching degree is output; When the association matching degree exceeds a preset matching threshold value, it is determined that the association condition is met.

6. The seat feedback adjustment system for supporting human posture acquisition according to claim 5, wherein The specific process of the feedback adjustment module calculating the posture deviation degree evaluation value is: Counting the number of associated pairs and the number of unassociated feature vectors in the first feature number sequence that meet the association condition; Counting the number of associated pairs and the number of unassociated feature vectors in the second feature number sequence that meet the association condition; According to the preset weight coefficient, the number of associated pairs and the number of unassociated feature vectors are weighted and fused to calculate.

7. The seat feedback adjustment system for supporting human posture acquisition according to claim 6, wherein The specific process of the feedback adjustment module generating the seat adjustment instruction is: Matching the posture deviation degree evaluation value with a plurality of preset adjustment level thresholds; According to the matching result, triggering the corresponding level of seat cushion height adjustment, backrest angle adjustment or waist support strength adjustment instruction.

8. The seat feedback adjustment system for supporting posture acquisition of a human body according to claim 7, wherein Further comprising: A resource scheduling module for monitoring the real-time operation load of the posture acquisition module, the partition processing module, the posture calibration module, the association verification module and the feedback adjustment module; A dynamic allocation module for dynamically allocating processing tasks to edge computing nodes through an index mapping mechanism according to the operation load data monitored by the resource scheduling module.

9. The seat feedback adjustment system for supporting posture acquisition of a human body according to claim 8, wherein The specific process of the dynamic allocation module is: Establishing the task processing capability index of each processing module and the resource state index of the edge computing node; Matching the task processing demand and the resource available state through a bidirectional index traversal algorithm; According to the matching result, the posture partition data processing task is real-time allocated to the target edge computing node.

10. The seat feedback adjustment system for supporting posture acquisition of a human body according to claim 9, wherein Further comprising a load balancing module for dynamically recycling or allocating the computing resources of each edge computing node through an elastic resource pool adjustment algorithm according to the task allocation record of the dynamic allocation module.

Images