A sitting posture intelligent monitoring method and system

Abstract

This invention provides a method and system for intelligent posture monitoring, relating to the field of data processing technology. The method includes: acquiring a user's posture dataset comprising multiple sets of multidimensional posture monitoring data and performing sliding window analysis to identify multiple local mutation windows and perform event fusion, determining multiple local mutation events in the multidimensional posture monitoring data; performing state segmentation on the multidimensional posture monitoring data, extracting multiple local steady-state events from the multidimensional posture monitoring data and determining multiple candidate postures; extracting multiple local steady-state features and multiple local mutation features, performing preference analysis on the multiple candidate postures to generate a user's posture preference list; acquiring a reference posture for the user and generating a posture deviation index for each candidate posture, optimizing the posture preference list to obtain a target posture matching list, and performing posture monitoring on the user based on the target posture matching list to generate posture monitoring analysis results. This invention achieves personalized intelligent posture monitoring.

Description

Technical Field

[0001] This invention relates to the field of data processing technology, and in particular to a sitting posture intelligent monitoring method and system. Background Technology

[0002] With changes in modern lifestyles, the monitoring and optimization of sitting posture has received increasing attention. Common posture monitoring methods are based on standard sitting posture assessments, focusing on whether the user conforms to a preset "ideal" sitting posture. By analyzing data such as posture angle, pressure distribution, and electromyography (EMG), these methods help users maintain a posture considered relatively healthy. However, for some users, maintaining this posture for extended periods can cause discomfort. Some posture monitoring methods often emphasize deviations from the standard posture, ignoring individual differences and comfort needs of users, and failing to fully consider the dynamic process of posture adjustment and the balance between comfort and health. While a standard sitting posture for extended periods may better meet the sitting health needs of the general population, it may cause fatigue for some users. Conversely, some postures that do not fully conform to the standard may provide greater comfort in the short term. Summary of the Invention

[0003] To address the aforementioned technical problems, this invention proposes a sitting posture intelligent monitoring method and system. By analyzing users' personalized comfort needs from sitting posture change data, analyzing users' adaptability to different sitting postures, and combining users' comfort and health needs, it provides more personalized sitting posture monitoring strategies, thereby improving the level of intelligent analysis in sitting posture monitoring.

[0004] The first aspect of this invention provides a method for intelligent monitoring of sitting posture, comprising:

[0005] The system acquires a user's sitting posture dataset, which includes multiple sets of multidimensional sitting posture monitoring data. It performs sliding window analysis on the multidimensional sitting posture monitoring data to identify multiple local mutation windows and their corresponding timestamps. It then performs event fusion on the multiple local mutation windows to determine multiple local mutation events for each set of multidimensional sitting posture monitoring data.

[0006] State segmentation is performed on multidimensional sitting posture monitoring data based on local mutation events, and multiple local steady-state events are extracted from the multidimensional sitting posture monitoring data. Multiple candidate sitting postures for multiple sets of multidimensional sitting posture monitoring data are determined based on multiple local steady-state events.

[0007] Local steady-state features of multiple local steady-state events and local mutation features of multiple local mutation events are extracted. Based on the multiple local steady-state features and local mutation features, a preference analysis is performed on multiple candidate sitting postures to generate a user's sitting posture preference list.

[0008] The system obtains reference sitting postures of users and performs posture deviation analysis on multiple candidate sitting postures to generate a posture deviation index for each candidate sitting posture. Based on the posture deviation index, the sitting posture preference list is optimized to obtain a target sitting posture matching list. Based on the target sitting posture matching list, the system monitors the user's sitting posture and generates sitting posture monitoring and analysis results.

[0009] Preferably, for local steady-state events and local abrupt events, it further includes:

[0010] In the process of sliding window analysis of multidimensional sitting posture monitoring data, the deviation features of each sitting posture monitoring item in each sliding window are extracted. If the deviation feature of any sitting posture monitoring item is greater than the preset deviation threshold of the sitting posture monitoring item, the sliding window is marked as a local mutation window. The nearest neighbor fusion of multiple local mutation windows is performed to generate multiple local mutation events corresponding to multiple local windows. Each local mutation event includes at least one local mutation window.

[0011] Local mutation events are removed from the multidimensional sitting posture monitoring data to obtain multiple candidate steady-state events. The distribution duration of each candidate steady-state event is determined and the event is filtered. Multiple candidate steady-state events with a distribution duration greater than a preset duration threshold are retained and marked as local steady-state events.

[0012] Preferably, the local steady-state features of multiple local steady-state events and the local mutation features of multiple local mutation events are extracted, including:

[0013] For local steady-state features, determine the multiple sets of multi-dimensional sitting posture monitoring data associated with each candidate sitting posture, construct the local sitting posture feature sequence of each local steady-state event in the multi-dimensional sitting posture monitoring data, and the target sitting posture feature sequence corresponding to the candidate sitting posture;

[0014] Multiple local sitting posture feature sequences are matched with the target sitting posture feature sequence to determine the initial steady-state event and multiple subordinate steady-state events in each group of multidimensional sitting posture monitoring data. The posture stability intensity parameter and posture deviation index corresponding to the initial steady-state event and multiple subordinate steady-state events are extracted to obtain the local steady-state features of the initial steady-state event and multiple subordinate steady-state events under the candidate sitting posture.

[0015] For local mutation features, based on the initial steady-state events in the multidimensional sitting posture monitoring data, we determine multiple subordinate mutation events in each set of multidimensional sitting posture monitoring data associated with the candidate sitting posture. We then extract the posture repair intensity parameter for each subordinate mutation event and the repair frequency index corresponding to the multiple subordinate mutation events in the multidimensional sitting posture monitoring data to obtain the local mutation features of multiple subordinate mutation events under the candidate sitting posture.

[0016] Preferably, a preference analysis is performed on multiple candidate sitting postures based on multiple local steady-state features and local abrupt change features to generate a user's sitting posture preference list, including:

[0017] The product of the posture stability strength parameter and the posture deviation index is calculated to generate the local steady-state index for each initial steady-state event and subordinate steady-state event. Multiple local steady-state indices under multidimensional sitting posture monitoring data are fused to obtain the global steady-state index of multidimensional sitting posture monitoring data. Based on the local sitting posture feature sequence and the target sitting posture feature sequence, the steady-state fusion weights corresponding to the multiple sets of multidimensional sitting posture monitoring data associated with each candidate sitting posture are calculated. The target steady-state index of each candidate sitting posture is calculated based on the steady-state fusion weights and the global steady-state index.

[0018] Based on the multiple local mutation features corresponding to the multiple sets of multidimensional sitting posture monitoring data associated with the candidate sitting posture, the global repair index corresponding to the multiple sets of multidimensional sitting posture monitoring data associated with the candidate sitting posture is calculated. The target repair index of each candidate sitting posture is calculated based on the steady-state fusion weight and the global repair index. The preference parameters of each candidate sitting posture are calculated based on the target steady-state index and the target repair index, and a user's sitting posture preference list for multiple candidate sitting postures is constructed.

[0019] Preferably, the sitting posture preference list is optimized based on the posture deviation index to obtain a target sitting posture matching list, and the user's sitting posture is monitored based on the target sitting posture matching list to generate sitting posture monitoring analysis results, including:

[0020] Based on the posture deviation index, the preference parameters of candidate sitting postures in the sitting posture preference list are corrected for posture deviation. The target matching parameters of each candidate sitting posture are obtained and a target sitting posture matching list is constructed. After collecting the user's real-time sitting posture monitoring data, the candidate sitting postures matched by the real-time sitting posture monitoring data are determined. The sitting posture monitoring analysis results are generated based on the candidate sitting postures and the target sitting posture matching list, including the sitting posture adaptation results and sitting posture optimization strategies based on the real-time sitting posture monitoring data.

[0021] Preferably, multiple candidate sitting postures are determined based on multiple local steady-state events regarding multiple sets of multidimensional sitting posture monitoring data, including:

[0022] Multiple local steady-state events are clustered based on the local sitting posture feature sequence of local steady-state events to generate multiple target sitting posture clusters for the user. Based on the multiple target sitting posture clusters, multiple candidate sitting postures are generated based on multiple sets of multi-dimensional sitting posture monitoring data.

[0023] A second aspect of the present invention provides a sitting posture intelligent monitoring system for implementing the above-mentioned sitting posture intelligent monitoring method, comprising:

[0024] The state mutation analysis module is used to acquire the user's sitting posture dataset, which includes multiple sets of multidimensional sitting posture monitoring data. It performs sliding window analysis on the multidimensional sitting posture monitoring data to identify multiple local mutation windows and their corresponding timestamps. It then performs event fusion on the multiple local mutation windows to determine multiple local mutation events for each set of multidimensional sitting posture monitoring data.

[0025] The candidate posture generation module is used to perform state segmentation on multidimensional sitting posture monitoring data based on local mutation events, extract multiple local steady-state events from the multidimensional sitting posture monitoring data, and determine multiple candidate sitting postures for multiple sets of multidimensional sitting posture monitoring data based on the multiple local steady-state events.

[0026] The sitting posture preference analysis module is used to extract the local steady-state features of multiple local steady-state events and the local mutation features of multiple local mutation events. Based on the multiple local steady-state features and local mutation features, the module performs preference analysis on multiple candidate sitting postures and generates a user's sitting posture preference list.

[0027] The posture monitoring and analysis module is used to obtain reference postures of users and perform posture deviation analysis on multiple candidate postures to generate a posture deviation index for each candidate posture. Based on the posture deviation index, the posture preference list is optimized to obtain a target posture matching list. Based on the target posture matching list, the user's posture is monitored to generate posture monitoring and analysis results.

[0028] The present invention has the following beneficial effects:

[0029] This invention conducts in-depth analysis of multi-dimensional sitting posture monitoring data, comprehensively considering the user's sitting posture stability, comfort, and adaptability. By extracting local steady-state features and abrupt change features, it identifies multiple candidate sitting postures preferred by the user, analyzes the user's adaptation status with different candidate sitting postures to represent the user's comfort status under different sitting postures, and combines a relatively standard reference sitting posture to analyze the user's adaptation with different candidate sitting postures from a health perspective. Finally, it generates a personalized sitting posture preference list by integrating the results from both health and comfort levels, realizing dynamic monitoring and optimization suggestions for the user's real-time sitting posture, thus improving the accuracy and intelligence of sitting posture monitoring. Attached Figure Description

[0030] Figure 1 This is a flowchart illustrating an exemplary sitting posture intelligent monitoring method according to an embodiment of the present invention.

[0031] Figure 2 This is a schematic diagram of the structure of an intelligent sitting posture monitoring system, which is an example of an embodiment of the present invention. Detailed Implementation

[0032] To enable those skilled in the art to better understand the technical solutions of this invention, the technical solutions of the embodiments of this invention will be clearly and completely described below with reference to the accompanying drawings. It should be understood that the specific embodiments described herein are only for explaining the invention and are not intended to limit the invention.

[0033] Please see Figure 1 The diagram illustrates a flowchart of an exemplary sitting posture intelligent monitoring method according to an embodiment of the present invention. The method specifically includes the following steps:

[0034] Step S01: Obtain the user's sitting posture dataset, which includes multiple sets of multidimensional sitting posture monitoring data. Perform sliding window analysis on the multidimensional sitting posture monitoring data to identify multiple local mutation windows and their corresponding timestamps. Perform event fusion on the multiple local mutation windows to determine multiple local mutation events for each set of multidimensional sitting posture monitoring data.

[0035] Specifically, multi-dimensional posture monitoring data includes, but is not limited to, user posture monitoring data, pressure monitoring data, and vital sign monitoring data on the smart chair. This data can be collected through sensors installed on smart devices or wearable devices. For example, pressure monitoring data can be collected through pressure sensors in the smart chair, user posture monitoring data can be collected through an inertial measurement unit, and vital sign monitoring data can be collected through devices such as electromyography, temperature, and blood pressure sensors. This represents the user's state in various dimensions while working or resting on the smart chair. Collecting relevant posture data of users on smart chairs through various smart devices is a mature existing technology, and this embodiment does not specifically limit the collection of multi-dimensional posture monitoring data. The collected data is organized into a user posture dataset, with each dataset representing the user's posture state within a time period.

[0036] Furthermore, continuous sitting posture data is segmented into events using sliding window analysis. During the sliding window's traversal of any set of multidimensional sitting posture monitoring data, each sliding window can be reasonably set according to the actual sitting posture analysis scenario, such as a few seconds or minutes. The deviation features of each sitting posture monitoring item within the sliding window are extracted; that is, the difference between the monitored value and the stable value used for reference. For example, the deviation feature is obtained by calculating the difference between the current spinal angle and the reference standard value. If the deviation feature of any sitting posture monitoring item is greater than a preset deviation threshold for that item, the sliding window is marked as a local mutation window. After determining multiple local windows in this way, nearest neighbor fusion is performed on these local mutation windows. For example, adjacent windows are merged to form multiple local mutation events. Each local mutation event includes at least one local mutation window. If a local mutation event includes more than one local mutation window, then any one of these local mutation windows has at least one adjacent local mutation window. Each local mutation event can be understood as a user's adjustment behavior to their sitting posture, reflecting the instability of the user's sitting posture and representing the user's adaptability to their posture.

[0037] Step S02: Perform state segmentation on the multidimensional sitting posture monitoring data based on local mutation events, extract multiple local steady-state events from the multidimensional sitting posture monitoring data, and determine multiple candidate sitting postures for multiple sets of multidimensional sitting posture monitoring data based on the multiple local steady-state events.

[0038] Specifically, after identifying multiple local abrupt change events, these events are removed from the multidimensional posture monitoring data to obtain multiple candidate steady-state events, indicating that the user maintains a relatively stable sitting posture after adjusting their posture. Considering that the candidate steady-state events obtained after the initial segmentation may be short-lived and not representative, the distribution duration of each candidate steady-state event is determined and the events are filtered. Multiple candidate steady-state events with a distribution duration or duration longer than a preset duration threshold are retained and marked as local steady-state events. Local steady-state events represent the stable sitting posture maintained by the user over a period of time, indicating the user's tendency or preference for a certain sitting posture.

[0039] For multiple local steady-state events, since they each represent a relatively stable state of the user while maintaining a certain sitting posture, several representative sitting postures can be identified by recognizing and summarizing the similarities between different sitting postures. In this process, the characteristic information related to sitting posture for each local steady-state event is first determined, and a local sitting posture feature sequence for each local steady-state event is constructed. This includes features such as posture angle features, pressure features, electromyographic features, and other physiological characteristics, such as the angle features of the spine, hips, and knees, the pressure features of various parts, and the intensity features of different muscle parts. The mean values ​​corresponding to different sitting posture monitoring items can be obtained statistically to complete the construction of the local sitting posture feature sequence.

[0040] Then, clustering algorithms such as K-means, DBSCAN, and hierarchical clustering are used to cluster the feature sequences of multiple local steady-state events, grouping similar sitting posture features into one category to generate multiple target sitting posture clusters. Each cluster represents a typical sitting posture preferred by the user. For example, using the DBSCAN algorithm for clustering, for the multiple target sitting posture clusters generated, based on the distance between different local sitting posture feature sequences and the cluster center, several representative local sitting posture feature sequences are determined as a candidate sitting posture. In this way, multiple candidate sitting postures are determined for multiple sets of multidimensional sitting posture monitoring data.

[0041] Step S03: Extract the local steady-state features of multiple local steady-state events and the local mutation features of multiple local mutation events. Based on the multiple local steady-state features and local mutation features, perform preference analysis on multiple candidate sitting postures to generate a user's sitting posture preference list.

[0042] Specifically, for the extraction of local steady-state features, firstly, multiple sets of multi-dimensional sitting posture monitoring data associated with each candidate sitting posture are identified. That is, within each target sitting posture cluster, multiple representative local sitting posture feature sequences are determined, and the local steady-state events to which these events belong are located. These multi-dimensional sitting posture monitoring data are then considered as the detection data associated with the candidate sitting posture. For each candidate sitting posture, the average of multiple representative local sitting posture feature sequences within the target sitting posture cluster is taken as the target sitting posture feature sequence representing the overall sitting posture state of that cluster.

[0043] Multiple local posture feature sequences are matched with the target posture feature sequence. Specifically, the similarity between sequences is calculated, for example, using cosine similarity. The local posture feature sequence with the highest similarity to the target posture feature sequence is designated as the initial steady-state event in the multi-dimensional posture monitoring data. Other local steady-state events located at the initial steady-state event are designated as subordinate steady-state events, representing the user reaching a representative posture and subsequent state changes based on that posture. Based on this, quantified local steady-state features are extracted from the posture maintenance level and the deviation level from the target state (i.e., the target posture feature sequence). This includes determining the maintenance duration of the initial steady-state event and multiple subordinate steady-state events, normalizing them to obtain the corresponding posture stability strength parameters, and calculating the similarity between different steady-state events and the target posture feature sequence to obtain the corresponding posture deviation index. This represents the user's specific maintenance state changes regarding these steady states.

[0044] The extraction process for local mutation features involves identifying multiple sets of multidimensional sitting posture monitoring data associated with the aforementioned candidate sitting postures, as well as the initial steady-state events within these data. Specifically, these are local mutation events occurring after the initial steady-state events in each set of multidimensional sitting posture monitoring data. Quantitative local mutation features are extracted from the perspectives of posture repair level and overall posture adjustment frequency. This includes determining the repair duration (event duration) of each subordinate mutation event, normalizing it to obtain the posture repair intensity parameter of the subordinate mutation event, and statistically analyzing the repair frequency data corresponding to multiple subordinate mutation events. After normalization, the repair frequency index corresponding to each set of multidimensional sitting posture monitoring data is obtained. This represents the user's self-local adjustment behavior regarding the candidate sitting posture.

[0045] After extracting multiple local steady-state features and local mutation features of different candidate sitting postures, preference analysis is performed on multiple candidate sitting postures based on these feature information, and finally a user's sitting posture preference list is generated according to different preference situations.

[0046] For the preference analysis process, the local steady-state features are processed as follows: First, the product of the posture stability strength parameter and the posture deviation index is calculated to generate the local steady-state index for each initial steady-state event and subordinate steady-state event. This considers the stability of the sitting posture and its deviation from the representative sitting posture. A higher local steady-state index generally indicates better stability, while a lower index indicates a lower level of user persistence and insufficient overall stability. Then, multiple local steady-state indices from the multi-dimensional sitting posture monitoring data are fused, including calculating the mean of multiple local steady-state indices to obtain the global steady-state index of the multi-dimensional sitting posture monitoring data. This index comprehensively characterizes the overall stability level of the user regarding the candidate sitting posture over a certain period. A higher global steady-state index indicates that the user can maintain this sitting posture for a longer period, suggesting that the user is likely more comfortable and has a higher preference for this posture.

[0047] For the specific stability levels corresponding to the multiple sets of multi-dimensional posture monitoring data associated with candidate postures, the steady-state fusion weights corresponding to the multiple sets of multi-dimensional posture monitoring data associated with each candidate posture are calculated based on the local posture feature sequence and the target posture feature sequence. Specifically, this can be generated after normalizing the similarity between the local posture feature sequence and the target posture feature sequence. The higher the similarity between sequences, the more reliable they are, indicating a higher degree of user preference for the candidate posture. By weighting and fusing the global steady-state index through the steady-state fusion weights, the target steady-state index of each candidate posture is finally calculated, representing the overall stability performance of the candidate posture in the multi-dimensional posture monitoring data. This can well reflect the user's stability in a certain posture, and a more stable posture indicates that the user can maintain that posture for a longer period of time.

[0048] For processing local steady-state features, based on the multiple local mutation features corresponding to the multiple sets of multi-dimensional sitting posture monitoring data associated with the candidate sitting posture, the global repair index corresponding to each set of multi-dimensional sitting posture monitoring data associated with the candidate sitting posture is calculated. In this embodiment, the product of the mean of multiple posture repair intensity parameters in the multi-dimensional sitting posture monitoring data and the repair frequency index corresponding to the multi-dimensional sitting posture monitoring data is used as the global repair index of the multi-dimensional sitting posture monitoring data. The longer the repair duration and the higher the repair frequency, the more self-adjustment phenomena the user exhibits in the candidate sitting posture, indicating the instability or maladaptability of the sitting posture. The global repair index is weighted and fused according to the steady-state fusion weight to calculate the target repair index for each candidate sitting posture, which represents the user's state recovery ability in the candidate sitting posture, as well as the ability to adjust and repair uncomfortable sitting postures. To a certain extent, it represents the instability of the sitting posture. A sitting posture with a large repair intensity and a high repair frequency usually means that the sitting posture is unstable for the user, and the user needs to frequently adjust the sitting posture to restore a comfortable state. Therefore, the target repair index can effectively measure the self-adjustment ability and adaptability of the sitting posture.

[0049] Finally, the preference parameters for each candidate sitting posture are calculated based on the target steady-state index and the target repair index. In this embodiment, the preference parameters are obtained by calculating the difference between the target steady-state index and the target repair index. The user's sitting posture preference list for multiple candidate sitting postures is then constructed by sorting the preference parameters. The target steady-state index focuses on the stability and health adaptability of the sitting posture, while the target repair index focuses on the comfort and self-adjustment ability of the sitting posture. Combining the two allows for a comprehensive and accurate assessment of the adaptability of each candidate sitting posture, obtaining data on the user's personalized needs. A larger preference parameter indicates a higher adaptability to the user, signifying that the user uses the candidate sitting posture more frequently in actual sitting conditions and requires less adjustment.

[0050] Step S04: Obtain the user's reference sitting posture and perform posture deviation analysis on multiple candidate sitting postures to generate a posture deviation index for each candidate sitting posture. Optimize the sitting posture preference list based on the posture deviation index to obtain a target sitting posture matching list. Perform sitting posture monitoring on the user based on the target sitting posture matching list to generate sitting posture monitoring and analysis results.

[0051] Specifically, the user's reference sitting posture can be a standard for healthy sitting posture set by professional medical staff based on the user's specific condition. For multiple candidate sitting postures extracted from a large amount of user monitoring data, a posture deviation analysis is performed on each candidate posture compared to the reference posture, calculating the corresponding posture deviation index. The posture deviation index reflects the degree of deviation between the candidate posture and the reference posture, including differences in dimensions such as spinal angle and pressure. The impact of differences in different dimensions can be reasonably set by professionals according to the actual situation. For example, if the user's spine is in an unhealthy state and needs timely repair, then the difference in spinal angle needs more attention. Ultimately, the posture deviation index characterizes the differences between different candidate sitting postures and the user's current sitting posture from a health perspective.

[0052] The target sitting posture matching list analyzes the matching status of different candidate sitting postures from the perspective of user's specific adaptability or personal preference. The posture deviation index assesses the adaptability of candidate sitting postures to users from a health perspective. Considering that a sitting posture considered relatively healthy may cause discomfort for users to maintain for a long time, it ignores individual differences and comfort needs of users. Therefore, it takes into account the user's actual comfort and health needs, and corrects the posture deviation of the preference parameters of candidate sitting postures in the sitting posture preference list based on the posture deviation index. The posture deviation index and preference parameters are combined to obtain the target matching parameter for each candidate sitting posture, where the target matching parameter = preference parameter × (1 - posture deviation index). Finally, the user's multiple candidate sitting postures are sorted according to the target matching parameter to obtain the target sitting posture matching list. The higher the target matching parameter, the more quantitative matching parameter after taking into account the user's comfort status and health needs. During the posture monitoring process using the target posture matching list, after collecting real-time posture monitoring data, candidate postures matching the real-time posture monitoring data are first determined. This involves matching the feature sequences corresponding to the real-time data with multiple candidate postures. After determining the most suitable candidate posture, the system doesn't solely assume a serious deviation in the user's posture from a health perspective. Instead, it considers the gradual process of posture adjustment, combining the user's comfort level in the current posture to comprehensively determine the adaptation result of the user's current posture. Based on this, and combined with the target posture matching list, several targets for localized posture optimization can be identified for the user's current state. An overly standard posture may result in poor user comfort; a phased adjustment approach can achieve a better transition effect, preventing low levels of discomfort from impacting the user's actual state, such as work or study efficiency. By improving the accuracy and intelligence of posture monitoring through this method, dynamic and real-time posture optimization suggestions can be provided to users. Combining posture improvement with comfort improvements can better achieve long-term health management and posture optimization.

[0053] Please see Figure 2This illustration shows a schematic diagram of an exemplary intelligent posture monitoring system according to an embodiment of the present invention. This system can be applied to smart chairs to achieve intelligent monitoring of the user's posture and assist the user in optimizing their posture for better health. Specifically, the system includes:

[0054] The state mutation analysis module is used to acquire the user's sitting posture dataset, which includes multiple sets of multidimensional sitting posture monitoring data. It performs sliding window analysis on the multidimensional sitting posture monitoring data to identify multiple local mutation windows and their corresponding timestamps. It then performs event fusion on the multiple local mutation windows to determine multiple local mutation events for each set of multidimensional sitting posture monitoring data.

[0055] The candidate posture generation module is used to perform state segmentation on multidimensional sitting posture monitoring data based on local mutation events, extract multiple local steady-state events from the multidimensional sitting posture monitoring data, and determine multiple candidate sitting postures for multiple sets of multidimensional sitting posture monitoring data based on the multiple local steady-state events.

[0056] The sitting posture preference analysis module is used to extract the local steady-state features of multiple local steady-state events and the local mutation features of multiple local mutation events. Based on the multiple local steady-state features and local mutation features, the module performs preference analysis on multiple candidate sitting postures and generates a user's sitting posture preference list.

[0057] The posture monitoring and analysis module is used to obtain reference postures of users and perform posture deviation analysis on multiple candidate postures to generate a posture deviation index for each candidate posture. Based on the posture deviation index, the posture preference list is optimized to obtain a target posture matching list. Based on the target posture matching list, the user's posture is monitored to generate posture monitoring and analysis results.

[0058] The modules in the aforementioned intelligent posture monitoring system employ the same technical means as the aforementioned intelligent posture monitoring method and can produce the same technical effects, which will not be elaborated here.

[0059] The above are merely specific embodiments of the present invention, enabling those skilled in the art to understand or implement the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art. Parts not described in detail in this specification are prior art known to those skilled in the art.

Claims

1. A method for intelligent monitoring of sitting posture, characterized in that, include: The system acquires a user's sitting posture dataset, which includes multiple sets of multidimensional sitting posture monitoring data. It performs sliding window analysis on the multidimensional sitting posture monitoring data to identify multiple local mutation windows and their corresponding timestamps. It then performs event fusion on the multiple local mutation windows to determine multiple local mutation events for each set of multidimensional sitting posture monitoring data. This includes extracting the deviation features of each sitting posture monitoring item within each sliding window. If the deviation feature of any sitting posture monitoring item is greater than the preset deviation threshold of the sitting posture monitoring item, the sliding window is marked as a local mutation window. The system then performs nearest neighbor fusion on the multiple local mutation windows to generate multiple local mutation events corresponding to the multiple local windows. Each local mutation event includes at least one local mutation window. The state segmentation of multidimensional sitting posture monitoring data is performed based on local mutation events, and multiple local steady-state events are extracted from the multidimensional sitting posture monitoring data. This includes removing local mutation events from the multidimensional sitting posture monitoring data to obtain multiple candidate steady-state events, determining the distribution duration of each candidate steady-state event and performing event filtering, retaining multiple candidate steady-state events with a distribution duration greater than a preset duration threshold and marking them as local steady-state events, and determining multiple candidate sitting postures for multiple sets of multidimensional sitting posture monitoring data based on multiple local steady-state events. Local steady-state features of multiple local steady-state events and local mutation features of multiple local mutation events are extracted. Based on the multiple local steady-state features and local mutation features, a preference analysis is performed on multiple candidate sitting postures to generate a user's sitting posture preference list. The system obtains reference sitting postures of users and performs posture deviation analysis on multiple candidate sitting postures to generate a posture deviation index for each candidate sitting posture. Based on the posture deviation index, the sitting posture preference list is optimized to obtain a target sitting posture matching list. Based on the target sitting posture matching list, the system monitors the user's sitting posture and generates sitting posture monitoring and analysis results.

2. The intelligent posture monitoring method according to claim 1, characterized in that, Local steady-state features of multiple local steady-state events and local catastrophe features of multiple local catastrophe events were extracted, including: For local steady-state features, determine the multiple sets of multi-dimensional sitting posture monitoring data associated with each candidate sitting posture, construct the local sitting posture feature sequence of each local steady-state event in the multi-dimensional sitting posture monitoring data, and the target sitting posture feature sequence corresponding to the candidate sitting posture; Multiple local sitting posture feature sequences are matched with the target sitting posture feature sequence to determine the initial steady-state event and multiple subordinate steady-state events in each group of multidimensional sitting posture monitoring data. The posture stability intensity parameters and posture deviation indices corresponding to the initial steady-state event and multiple subordinate steady-state events are extracted, including determining the maintenance duration corresponding to the initial steady-state event and multiple subordinate steady-state events. After normalization, the corresponding posture stability intensity parameters are obtained, and the similarity between different steady-state events and the target sitting posture feature sequence is calculated to obtain the corresponding posture deviation index. The local steady-state features of the initial steady-state event and multiple subordinate steady-state events under the candidate sitting posture are obtained. For local mutation features, based on the initial steady-state events in the multidimensional sitting posture monitoring data, multiple subordinate mutation events are determined in each set of multidimensional sitting posture monitoring data associated with the candidate sitting posture. The posture repair intensity parameter of each subordinate mutation event is extracted, including determining the event duration of each subordinate mutation event and obtaining the posture repair intensity parameter of the subordinate mutation event after normalization. The repair frequency index corresponding to multiple subordinate mutation events in the multidimensional sitting posture monitoring data is also obtained, including statistically analyzing the repair frequency data corresponding to multiple subordinate mutation events and obtaining the repair frequency index corresponding to each set of multidimensional sitting posture monitoring data after normalization. Thus, the local mutation features of multiple subordinate mutation events under the candidate sitting posture are obtained.

3. The intelligent posture monitoring method according to claim 2, characterized in that, Based on multiple local steady-state features and local abrupt change features, a preference analysis is performed on multiple candidate sitting postures to generate a user's sitting posture preference list, including: The product of the posture stability strength parameter and the posture deviation index is calculated to generate the local steady-state index for each initial steady-state event and subordinate steady-state event. Multiple local steady-state indices under multidimensional sitting posture monitoring data are fused to obtain the global steady-state index of multidimensional sitting posture monitoring data. Based on the local sitting posture feature sequence and the target sitting posture feature sequence, the steady-state fusion weights corresponding to the multiple sets of multidimensional sitting posture monitoring data associated with each candidate sitting posture are calculated. The steady-state fusion weights are generated by normalizing the similarity between the local sitting posture feature sequence and the target sitting posture feature sequence. The target steady-state index of each candidate sitting posture is calculated based on the steady-state fusion weights and the global steady-state index, including weighted fusion of the global steady-state index through steady-state fusion weights. Based on the multiple local mutation features corresponding to the multiple sets of multidimensional sitting posture monitoring data associated with the candidate sitting posture, the global repair index corresponding to the multiple sets of multidimensional sitting posture monitoring data associated with the candidate sitting posture is calculated. This includes multiplying the mean of multiple posture repair intensity parameters in the multidimensional sitting posture monitoring data with the repair frequency index corresponding to the multidimensional sitting posture monitoring data as the global repair index of the multidimensional sitting posture monitoring data. The target repair index of each candidate sitting posture is calculated based on the steady-state fusion weight and the global repair index. This includes weighting and fusing the global repair index according to the steady-state fusion weight to calculate the target repair index of each candidate sitting posture. The preference parameters of each candidate sitting posture are calculated based on the target steady-state index and the target repair index. The preference parameters are obtained by calculating the difference between the target steady-state index and the target repair index. This constructs a user's sitting posture preference list for multiple candidate sitting postures.

4. The intelligent posture monitoring method according to claim 3, characterized in that, The sitting posture preference list is optimized based on the posture deviation index to obtain a target sitting posture matching list. Based on this target sitting posture matching list, user sitting posture is monitored to generate sitting posture monitoring and analysis results, including: Based on the posture deviation index, the preference parameters of candidate sitting postures in the sitting posture preference list are corrected for posture deviation. The target matching parameters of each candidate sitting posture are obtained and a target sitting posture matching list is constructed. This includes fusing the posture deviation index and preference parameters to obtain the target matching parameters of each candidate sitting posture, where the target matching parameter = preference parameter × (1 − posture deviation index). The multiple candidate sitting postures of the user are sorted according to the target matching parameters to obtain the target sitting posture matching list. After collecting the user's real-time sitting posture monitoring data, the candidate sitting postures matched by the real-time sitting posture monitoring data are determined. The sitting posture monitoring analysis results are generated based on the candidate sitting postures and the target sitting posture matching list, including the sitting posture adaptation results and sitting posture optimization strategies based on the real-time sitting posture monitoring data.

5. The intelligent posture monitoring method according to claim 4, characterized in that, Based on multiple local steady-state events, several candidate sitting postures were determined from multiple sets of multidimensional sitting posture monitoring data, including: Multiple local steady-state events are clustered based on the local sitting posture feature sequence of local steady-state events to generate multiple target sitting posture clusters for the user. Based on the multiple target sitting posture clusters, multiple candidate sitting postures are generated based on multiple sets of multi-dimensional sitting posture monitoring data.

6. A sitting posture intelligent monitoring system, characterized in that, The system is used to implement the intelligent posture monitoring method according to any one of claims 1-5, comprising: The state mutation analysis module is used to acquire the user's sitting posture dataset, which includes multiple sets of multidimensional sitting posture monitoring data. It performs sliding window analysis on the multidimensional sitting posture monitoring data to identify multiple local mutation windows and their corresponding timestamps. It performs event fusion on the multiple local mutation windows to determine multiple local mutation events for each set of multidimensional sitting posture monitoring data. This includes extracting the deviation features of each sitting posture monitoring item within each sliding window. If the deviation feature of any sitting posture monitoring item is greater than the preset deviation threshold of the sitting posture monitoring item, the sliding window is marked as a local mutation window. It performs nearest neighbor fusion on multiple local mutation windows to generate multiple local mutation events corresponding to multiple local windows. Each local mutation event includes at least one local mutation window. The candidate posture generation module is used to perform state segmentation on multidimensional sitting posture monitoring data based on local mutation events, extract multiple local steady-state events from the multidimensional sitting posture monitoring data, including removing local mutation events from the multidimensional sitting posture monitoring data to obtain multiple candidate steady-state events, determining the distribution duration of each candidate steady-state event and performing event filtering, retaining multiple candidate steady-state events with a distribution duration greater than a preset duration threshold and marking them as local steady-state events, and determining multiple candidate sitting postures for multiple sets of multidimensional sitting posture monitoring data based on multiple local steady-state events; The sitting posture preference analysis module is used to extract the local steady-state features of multiple local steady-state events and the local mutation features of multiple local mutation events. Based on the multiple local steady-state features and local mutation features, the module performs preference analysis on multiple candidate sitting postures and generates a user's sitting posture preference list. The posture monitoring and analysis module is used to obtain reference postures of users and perform posture deviation analysis on multiple candidate postures to generate a posture deviation index for each candidate posture. Based on the posture deviation index, the posture preference list is optimized to obtain a target posture matching list. Based on the target posture matching list, the user's posture is monitored to generate posture monitoring and analysis results.

Images