Sitting posture self-adaptive adjusting method and system based on multi-modal data and seat

Abstract

The invention relates to the field of seat design, in particular to design optimization for only adjusting a seat supporting mode according to the sitting posture of a user. The invention provides a sitting posture adaptive adjustment method based on multi-modal data, and the method comprises the steps: collecting a pressure distribution receipt through a multi-dimensional pressure array sensor, and collecting trunk attitude angle data through an inertial measurement unit; based on the pressure distribution data and the trunk attitude angle data, determining a current posture of the user; and based on the current posture of the user, executing a posture strategy corresponding to the current posture according to the predicted strategy rule. According to the sitting posture self-adaptive adjusting method and system based on the multi-modal data and the seat, the actual sitting posture of the user can be effectively recognized with high precision through the multi-modal collected data, and the accuracy and fitness of the supporting effect of the supporting body are improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of seat design, in particular to the design optimization of seat support mode adjustment according to user sitting posture. BACKGROUND

[0002] A seat is a device that provides sitting support for a person, aiming to improve comfort, maintain a healthy posture, and support work and rest. The support body is the core component of the seat, such as the backrest and the lumbar support. Its core function is to effectively distribute the load generated by the body weight through reasonable structural design, and to provide stable support for the human body to reduce the pressure on the muscle and skeletal system, especially the load on the lumbar spine.

[0003] In the current mainstream products, the adjustment mechanism of the support body mainly relies on manual adjustment and single sensor triggering. Among them, manual adjustment is for users to adjust the height, depth of the lumbar support or the inclination of the seat cushion through physical controls such as mechanical switches, knobs or handles according to their own subjective feelings. This adjustment is a one-time, static setting and cannot be automatically adjusted during use according to the dynamic changes in user posture.

[0004] The application of a single sensor is also relatively simple. For example, a single pressure sensor is placed inside the seat cushion or lumbar support. Its logic is usually a binary judgment. When pressure is detected, a fixed support mode is activated, and when the pressure disappears, it is turned off. In fact, users have different body types and postures are variable. Under this scheme, the user's sitting posture cannot be accurately obtained, that is, it cannot accurately determine whether the user is sitting upright, leaning back, leaning forward or leaning sideways. The comfort of the support effect of the support body based on this identification is also poor. SUMMARY

[0005] The purpose of the present application is to provide a sitting posture adaptive adjustment method and system based on multi-modal data and a seat. Through the collection of multi-modal data, the actual sitting posture of the user can be effectively and accurately identified, and the accuracy and adaptability of the support effect of the support body can be improved.

[0006] The present application provides a sitting posture adaptive adjustment method based on multi-modal data, the method comprising:

[0007] Collecting pressure distribution records through a multi-dimensional pressure array sensor and collecting torso posture angle data through an inertial measurement unit;

[0008] Determining the current posture of the user based on the pressure distribution data and the torso posture angle data;

[0009] Based on the current posture of the user, executing the posture strategy corresponding to the current posture according to the predicted strategy rule.

[0010] As a preferred embodiment of the present application, during the process of collecting the torso posture angle data by the inertial measurement unit, a Kalman filter sensor fusion algorithm is used to fuse the data of the three-dimensional gyroscope and the three-axis accelerometer to obtain the torso posture angle data.

[0011] As a preferred embodiment of the present application, based on the pressure distribution data and the torso posture angle data, the current posture of the user is determined, including:

[0012] The pressure distribution data and the torso posture angle data are filtered, the pressure distribution data is subjected to median filtering, and the torso posture angle data is subjected to sliding average filtering.

[0013] As a preferred embodiment of the present application, based on the pressure distribution data and the torso posture angle data, the current posture of the user is determined, including:

[0014] According to the pressure distribution data and the torso posture angle data, a pressure topology feature vector V_pressure and a posture feature vector V_pose are obtained respectively, and the two are fused to obtain a joint feature vector V_fused, and the current posture of the user is determined according to the joint feature vector V_fused.

[0015] As a preferred embodiment of the present application, based on the joint feature vector V_fused, an n-dimensional logical vector logits is obtained by a sofmax classifier, n being a preset number of postures;

[0016] The softmax classifier converts the logical vector logits into a probability distribution, P(i) = exp(l_i) / (exp(l1) + exp(l2) +... + exp(ln));

[0017] The class with the highest probability value is selected as the current posture of the user.

[0018] As a preferred embodiment of the present application, a micro-pump system and a plurality of air bags are installed in the seat support body, and the posture strategy includes pressure control of the micro-pump system to adjust the air bag pressure.

[0019] As a preferred embodiment of the present application, during the process of determining the current posture of the user based on the pressure distribution data and the torso posture angle data, a confidence score of the current posture of the user is obtained, and when the confidence score corresponding to the posture continuously exceeds a preset threshold value for a preset time, the next step of the sitting posture execution action is performed.

[0020] As a preferred embodiment of the present application, when the difference between the confidence scores corresponding to the two postures with the highest confidence scores is lower than a preset value, it is judged that this is a posture transition state, and a preset transition execution scheme is executed

[0021] A sitting posture adaptive adjustment system based on multi-modal data, the system comprising:

[0022] A data acquisition module configured to collect pressure distribution data through a multi-dimensional pressure array sensor and torso posture angle data through an inertial measurement unit;

[0023] A posture determination module configured to determine a current posture of a user based on the pressure distribution data and the torso posture angle data;

[0024] A strategy execution module configured to execute a posture strategy corresponding to the current posture according to a predicted strategy rule based on the current posture of the user.

[0025] A seat comprising a main frame, a support body provided on the main frame, a multi-dimensional pressure array sensor and an inertial measurement unit arranged on the support body, and a memory, wherein the memory stores a computer program, and the computer program is executed by a processor to implement the method of any one of claims 1-8.

[0026] Advantages:

[0027] 1. By fusing pressure distribution and torso posture angle multi-modal data and using a hybrid neural network model for feature extraction and classification, the system can accurately identify a variety of typical sitting postures.

[0028] 2. Based on the identified sitting posture, the system dynamically adjusts the pressure of the lumbar support airbag and uses an L3-L5 force balancing algorithm for closed-loop fine-tuning to achieve millimeter-level accurate support, significantly improving comfort and effectively reducing lumbar load, and the support effect is optimized in real time according to the change in posture.

[0029] 3. The confidence mechanism is introduced as the basis for execution judgment, and only when the confidence is continuously high will the action be taken, and a conservative strategy is executed for the transition posture, effectively avoiding false triggering and frequent adjustment, making the system response smoother and more stable, and the user experience more natural.

[0030] 4. The system links posture recognition with support adjustment, massage, hot compress and other functions, triggers a complete set of coordinated strategies according to a specific posture, realizes integrated intelligent service from perception to execution, and breaks through the limitations of isolated operation of each functional module.

[0031] 5. The closed-loop control mechanism of perception-decision-execution-feedback enables the system to continuously monitor the pressure distribution and dynamically adjust the airbag, ensuring that the user's lumbar spine is actively, continuously and evenly supported in any sitting posture, and truly realizing personalized adaptation. BRIEF DESCRIPTION OF DRAWINGS

[0032] Figure 1A flowchart of some embodiments of the present specification is shown;

[0033] Figure 2 A flowchart of the specific details of determining the current posture of the user of some embodiments of the present specification is shown;

[0034] Figure 3 A structural schematic diagram of an electronic device of some embodiments of the present specification is shown;

[0035] Figure 4 A structural schematic diagram of a system of some embodiments of the present specification is shown. DETAILED DESCRIPTION

[0036] The present application will be further described in conjunction with the drawings.

[0037] The technical solutions in the embodiments of the present specification will be clearly and completely described below in conjunction with the drawings in the embodiments of the present specification.

[0038] The terms "first", "second", "third", etc. in the specification and claims of the present specification and the above drawings are used to distinguish different objects, and are not used to describe a specific order. In addition, the terms "include" and "have" and any variations thereof are intended to cover non-exclusive inclusion. For example, a process, method, system, product or device that includes a series of steps or units is not limited to the listed steps or units, but can optionally include steps or units not listed, or can optionally include other steps or units inherent to the process, method, product or device.

[0039] The following description provides examples and does not limit the scope, applicability or examples set forth in the claims. Changes can be made to the functions and arrangements of the described elements without departing from the scope of the present specification. Various examples can appropriately omit, replace or add various processes or components. For example, the described methods can be performed in a different order from the described order, and various steps can be added, omitted or combined. In addition, features described with respect to some examples can be combined into other examples.

[0040] Embodiments of the present disclosure propose a sitting posture adaptive adjustment method based on multi-modal data, Figure 1 A flowchart of the sitting posture adaptive adjustment method based on multi-modal data of some embodiments of the present specification is shown, and below, the specific operation steps of the method 100 are exemplarily explained.

[0041] In block 101, the method 100 collects user data. The user data here contains at least two aspects, which are pressure distribution data and torso posture angle data, respectively.

[0042] In the support body of the seat, a multi-dimensional pressure array sensor is arranged. The sensor is a sensor in which a large number of micro pressure-sensitive units are integrated in a grid array on a flexible film. It is usually composed of a special piezoresistive material sandwiched between two layers of flexible circuit board FPC. The resistance of the piezoresistive material changes linearly with the size of the pressure applied.

[0043] The pressure distribution data collected by the multi-dimensional pressure array sensor is a two-dimensional matrix containing pressure values and position information. The value of each point represents the relative pressure intensity at that coordinate.

[0044] In the support body of the seat, an inertial measurement unit IMU is also arranged. The IMU is a micro-electro-mechanical system chip that integrates a multi-axis sensor module. A typical IMU contains two key sensors: a three-axis accelerometer that measures linear acceleration including gravity. When the device is stationary, it mainly senses the direction of gravitational acceleration, so it can determine the inclination angle of the device relative to the gravity line. A three-axis gyroscope that measures the angular velocity of rotation around the X, Y, and Z coordinate axes. By integrating the angular velocity over time, the angular change can be calculated.

[0045] In this embodiment, the torso posture angle data is not simply obtained from the angular change of the three-axis gyroscope, but a sensor fusion algorithm such as Kalman filtering is used to fuse the data of the three-dimensional gyroscope and the three-axis accelerometer to obtain the final torso posture angle data.

[0046] This is because the gyroscope can measure angular velocity very sensitively. By integrating the angular velocity, the angular change can be calculated. For example, if a 10-degree angular velocity per second is detected for 0.1 seconds, the angle changes by 1 degree. It has very high accuracy in the short term. However, the gyroscope has a zero-point drift error. The accelerometer provides an absolute gravity reference and has no long-term drift problem. Therefore, the accelerometer continuously calculates an absolute angle based on the direction of gravity, and uses the gravity reference to gradually correct the drift of the gyroscope.

[0047] The final output of the hardware system is the stable three-dimensional posture angle data after fusion calculation.

[0048] The final data is usually presented in the form of a set of three Euler angles, which directly describes the orientation of the support body in three-dimensional space. For example, {pitch: -15.5, roll: 1.2, yaw: 0.3}. This indicates that the current posture of the support body is tilted forward by 15.5 degrees, slightly tilted to the right by 1.2 degrees, and slightly twisted to the right by 0.3 degrees.

[0049] In box 102, method 100 determines the user's current posture based on the pressure distribution data and torso posture angle data described above.

[0050] like Figure 2 As shown, Figure 2 The specific sub-steps of this process are determined for the posture.

[0051] Within box 1021, the data is preprocessed, especially filtered.

[0052] The filtering methods for pressure distribution data and drive attitude angle data differ here. For the former, median filtering can be used. This is because pressure data is easily affected by slight vibrations, clothing friction, or electronic noise from the sensor itself. In this sub-step, the values ​​of a pressure point and its neighboring points, such as a 3x3 window, are sorted, and the median is taken as the new output value for that point.

[0053] For example, a pressure value of 90 before filtering exceeds the normal pressure value. Taking this value as the center point, take 3x3 data points: [52, 50, 48], [55, 90, 49], [51, 53, 50]. After sorting, the result is: [48, 49, 50, 50, 51, 52, 53, 55, 90]. The median is 51. Replace the noise value 90 with 51.

[0054] The torso attitude angle data from the inertial measurement unit (IMU) can be processed using a moving average filter. This filtering method takes a continuous data stream, takes data points from the current moment and several previous moments, calculates their arithmetic mean, and uses this as the output for the current moment. This effectively smooths out high-frequency random noise, making the data curve smoother.

[0055] In box 1022, a neural network is used to extract features from the pressure distribution data and the torso posture angle data.

[0056] Ideally, two different neural network models are used for parallel processing.

[0057] The pressure distribution data is processed using a graph neural network (GNN). After multiple layers of graph convolution, each node contains topological information about its local neighborhood. Subsequently, through global pooling operations, such as averaging the features of all nodes, the information of the entire graph is aggregated into a fixed-length pressure topological feature vector, V_pressure.

[0058] The torso pose angle data is processed using a feedforward neural network (FNN), which supports connecting the three pose angles [pitch, roll, yaw] into a 3-dimensional feature vector, namely the pose feature vector, V_pose.

[0059] Specifically, in box 1023, the pressure topology feature vector, V_pressure, and the attitude feature vector, V_pose, from the previous step are fused together to form a joint feature vector, V_fused.

[0060] The process then proceeds to box 1024, which is the last step in box 102, namely determining the user's pose based on the joint feature vector V_fused described above.

[0061] The main body of this stage typically consists of one or more fully connected layers and a Softmax classifier.

[0062] For example, the role of a fully connected layer is to perform final feature transformation and dimensionality compression, learning a more complex non-linear relationship between joint features and pose category. Its presence is an optimization step, but it may or may not be present.

[0063] The softmax layer is the core component of this stage, responsible for making the final classification and recognition decision. It maps the input vector through a linear layer to a dimension equal to the number of categories. For example, in this embodiment, if the user's pose is set to a total of 6 possibilities, the linear layer will convert the input into a 6-dimensional logistic vector, i.e., logits.

[0064] Logits = W2 * Hidden + b2

[0065] W2 is the weight matrix, and b2 is the bias. Logits is a 6-dimensional vector, such as [l1, l2, l3, l4, l5, l6]. The magnitude of each value roughly represents the recognition strength of the input data belonging to the corresponding category. Softmax then transforms the logits into a probability distribution.

[0066] The probability of the i-th category is P(i) = exp(l_i) / (exp(l1) + exp(l2) + ... + exp(l6)).

[0067] This operation amplifies the probability corresponding to the maximum logits value through an exponential function, making the difference between the highest and second-highest probabilities more significant. Normalization is then performed to ensure that the sum of the probabilities of all categories is 1. The argmax function is then used to select the category with the highest probability as the final recognition result.

[0068] Predicted_Class = argmax([P1, P2, P3, P4, P5, P6])

[0069] For example, the six posture categories are: 0: standard sitting, 1: leaning forward while working, 2: leaning back while resting, 3: leaning to the left, 4: leaning to the right, and 5: bending forward.

[0070] The data from box 1023 is multidimensional, for example, a 192-dimensional joint feature vector V_fused. After passing through a fully connected layer and Softmax, it yields 6 probability values:

[0071] P = [0.02, 0.95, 0.01, 0.01, 0.005, 0.005].

[0072] The probability of sitting upright is 2%, leaning forward while working is 95%, leaning back while resting is 1%, leaning to the left is 1%, leaning to the right is 0.5%, and bending forward is 0.5%. The forward-leaning posture corresponding to 95% of the maximum values ​​is the posture identified in this step.

[0073] At this point, box 102 ends, outputting not only the final recognized posture (forward-leaning office) but also the corresponding confidence score, i.e., the maximum probability value, 0.95.

[0074] Next, proceed to box 103, which is an execution step in this embodiment. The designers have already implemented strategies corresponding to each posture. For example, in a forward-leaning office posture: ensure the strongest support for the lower lumbar spine, namely the L3-L5 region. Preferably, acupressure and local heat application can also be activated, both focusing on the L3-L5 region. In a backward-leaning resting posture: allow the entire back to bear the force evenly, with the lumbar support point appropriately moved upward and softened. Preferably, the massage area is the entire back, and the lumbar support heat application is also applied to the entire back.

[0075] The above strategies have already been stored in the storage. Now, we only need to query and retrieve them to execute the forward-leaning office execution strategy.

[0076] For example, there is a design here for dynamic monitoring and regulation of the execution process.

[0077] Specifically, the support structure, such as the lumbar support, incorporates a distributed micro-air pump system containing numerous airbags. As mentioned above, the seat's support structure also includes a multi-dimensional pressure array sensor. At this point, a forward-leaning office posture has been identified, requiring enhanced support for the L3-L5 lumbar region. The system has a built-in optimization model, such as an optimization algorithm model based on the proportional-integral-derivative control principle. The system continuously receives real-time data streams from the multi-dimensional pressure array sensor. This data constitutes a dynamic two-dimensional pressure distribution map of the lumbar-lumbar support contact interface. The algorithm analyzes the current pressure distribution map in real time, calculating the average or peak pressure values ​​at the corresponding anatomical positions of L3, L4, and L5.

[0078] The system uses a built-in proportional-integral-derivative (PI-DI) control algorithm to calculate the airbag pressure adjustment required to achieve pressure equilibrium. For example, if the pressure in region L4 is detected to be significantly higher than that in regions L3 and L5, the algorithm calculates the pressure increase for the corresponding airbags in regions L3 and L5 and sends this information to the distributed micro-pump system described above. The air pumps, based on the instructions, inflate or deflate the independent airbag units inside the lumbar support corresponding to specific lumbar vertebral segments. This adjustment is zoned and pressure-divided, achieving millimeter-level dynamic reconstruction of the support contour.

[0079] After the action is executed, the pressure array sensors immediately detect the new pressure distribution. The new pressure data is fed back to the system, and the algorithm compares it with the equilibrium target to determine whether the optimal state has been reached. If not, the previous step is repeated for the next round of fine-tuning.

[0080] In some embodiments, after the completion of the above-mentioned box 102, there is an additional data design for the output recognition result and the corresponding confidence score.

[0081] In some embodiments, the confidence score has a time-based discrimination function. The next step of the sitting posture execution action is only performed when the confidence score of the sitting posture is higher than a threshold and the time above the threshold exceeds a preset time.

[0082] For example, if a user simply moves their body, the system recognizes a 92% confidence score for a backward leaning posture, but only maintains this for one second. Instead of executing any action at this point, the system waits until the confidence score exceeds a preset time threshold, such as more than three seconds, before confirming it as a backward leaning posture and proceeding to the next seating action. This avoids erroneous or overly sensitive responses from the chair, thus improving the user experience.

[0083] In some embodiments, the confidence score has a transition identification function. When the difference between the first confidence score and the second confidence score is lower than a preset threshold, it is determined that the posture transition state is in effect, and a preset transition execution scheme is executed.

[0084] For example, P = [0.55, 0.45, 0.02, 0.02, 0.03, 0.03]. The confidence scores for standard upright sitting and leaning forward while working are 0.55 and 0.45 respectively, with a difference of 0.1, which is less than the preset threshold of 0.2. At this point, the system recognizes it as a transitional state from leaning forward to standard upright sitting. The transition execution plan is then implemented. This transition execution plan is preset by the designer; for example, it may only perform basic air pressure maintenance without switching to massage or heat therapy modes; or, it may read sensor data more frequently and make small, smooth adjustments until the confidence level rises above the threshold; or, it may maintain the current support and heat therapy modes to avoid abrupt changes for the user until the confidence level rises above the threshold.

[0085] In the above embodiments, implementation can be achieved, in whole or in part, through software, hardware, firmware, or any combination thereof. When implemented in software, it can be implemented, in whole or in part, as a computer program product. The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, all or part of the processes or functions described in the embodiments of this specification are generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in or transmitted through a computer-readable storage medium. The computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired (e.g., coaxial cable, fiber optic, Digital Subscriber Line (DSL)) or wireless (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium can be any available medium accessible to a computer or a data storage device such as a server or data center that integrates one or more available media. The available media may be magnetic media (e.g., floppy disks, hard disks, magnetic tapes), optical media (e.g., Digital Versatile Discs (DVDs)), or semiconductor media (e.g., Solid State Disks (SSDs)).

[0086] Figure 4This diagram illustrates the structure of a multimodal data-based adaptive posture adjustment system 400 according to some embodiments of this disclosure. The various embodiments in this specification are described in a progressive manner; similar or identical parts between embodiments can be referred to mutually. Each embodiment focuses on describing the differences from other embodiments. In particular, the device embodiments are basically similar to the method embodiments, so the description is relatively simple; relevant parts can be referred to the descriptions of the method embodiments. Figure 4 As shown, system 400 includes:

[0087] The data acquisition module 401 is configured to collect pressure distribution data through a multi-dimensional pressure array sensor and torso attitude angle data through an inertial measurement unit.

[0088] The posture determination module 402 is configured to determine the user's current posture based on the pressure distribution data and the torso posture angle data;

[0089] The strategy execution module 403 is configured to execute the posture strategy corresponding to the current posture based on the user's current posture and according to the predicted strategy rules.

[0090] Figure 3 A block diagram of an electronic device 300 that can implement various embodiments of the present disclosure is shown. For example... Figure 3 As shown, device 300 includes a processor 301, which can perform various appropriate actions and processes based on computer program instructions loaded into random access memory (RAM) 303 according to computer program instructions stored in read-only memory (ROM) 302. RAM 303 may also store various programs and data required for the operation of device 300. The processor 301, ROM 302, and RAM 303 are interconnected via bus 304. Input / output (I / O) interface 305 is also connected to bus 304.

[0091] The various processes and procedures described above, such as method 100, can be executed by processor 301. For example, in some embodiments, method 100 may be implemented as a software program tangibly contained in a machine-readable medium. In some embodiments, part or all of the software program may be loaded and / or installed on device 300 via ROM 302. When the software program is loaded into RAM 303 and executed by processor 301, one or more actions of method 100 described above may be performed.

[0092] The program code used to implement the methods of this disclosure may be written in any combination of one or more programming languages. This program code may be provided to a processor or controller of a general-purpose computer, special-purpose computer, or other programmable data processing apparatus, such that when executed by the processor or controller, the program code causes the functions / operations specified in the flowcharts and / or block diagrams to be implemented. The program code may be executed entirely on a machine, partially on a machine, as a standalone software package partially on a machine and partially on a remote machine, or entirely on a remote machine or server.

[0093] In the context of this disclosure, a machine-readable medium can be a tangible medium that may contain or store a program for use by or in conjunction with an instruction execution system, apparatus, or device. A machine-readable medium can be a machine-readable signal medium or a machine-readable storage medium. Machine-readable media can be, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatus, or devices, or any suitable combination of the foregoing. More specific examples of machine-readable storage media include electrical connections based on one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination of the foregoing. Furthermore, although operations are depicted in a specific order, this should be understood as requiring that such operations be performed in the specific order shown or in sequential order, or requiring that all illustrated operations be performed to achieve the desired result. In certain environments, multitasking and parallel processing may be advantageous. Similarly, while several specific implementation details are included in the foregoing discussion, these should not be construed as limiting the scope of this disclosure. Certain features described in the context of individual embodiments may also be implemented in combination in a single implementation. Conversely, various features described in the context of a single implementation may also be implemented individually or in any suitable sub-combination in multiple implementations.

[0094] Although the subject matter has been described using language specific to structural features and / or methodological logic, it should be understood that the subject matter defined in the appended claims is not necessarily limited to the specific features or actions described above. Rather, the specific features and actions described above are merely illustrative examples of implementing the claims.

Claims

1. A method for adaptive adjustment of sitting posture based on multi-modal data, characterized in that, The method comprises: collecting a pressure distribution receipt through a multi-dimensional pressure array sensor, and collecting torso posture angle data through an inertial measurement unit; determining a current posture of a user based on the pressure distribution data and the torso posture angle data; based on the current posture of the user, performing a posture strategy corresponding to the current posture according to a predicted strategy rule.

2. The method for adaptive adjustment of sitting posture based on multi-modal data according to claim 1, characterized in that: In the process of collecting torso posture angle data through an inertial measurement unit, a Kalman filter sensor fusion algorithm is used to fuse the data of a three-dimensional gyroscope and a three-axis accelerometer to obtain the torso posture angle data.

3. The method for adaptive adjustment of sitting posture based on multi-modal data according to claim 1, characterized in that: Based on the pressure distribution data and the torso posture angle data, determining a current posture of a user, comprises: filtering the pressure distribution data and the torso posture angle data, performing median filtering on the pressure distribution data, and performing sliding average filtering on the torso posture angle data.

4. The method for adaptive adjustment of sitting posture based on multi-modal data according to claim 1, characterized in that: Based on the pressure distribution data and the torso posture angle data, determining a current posture of a user, comprises: obtaining a pressure topology feature vector V_pressure and a posture feature vector V_pose from the pressure distribution data and the torso posture angle data respectively, fusing the two to obtain a joint feature vector V_fused, and determining a current posture of a user according to the joint feature vector V_fused.

5. The method for adaptive adjustment of sitting posture based on multi-modal data according to claim 4, characterized in that: Based on the joint feature vector V_fused, an n-dimensional logic vector logits is obtained through a sofmax classifier, where n is a preset number of postures; The softmax classifier converts the logic vector logits into a probability distribution, P(i) = exp(l_i) / (exp(l1) + exp(l2) +... + exp(ln)); selecting the class with the highest probability value as the current posture of the user.

6. The method for adaptive adjustment of sitting posture based on multi-modal data according to claim 1, characterized in that: A micro-pump system and multiple air bags are installed in the seat support body, and the posture strategy comprises pressure control of the micro-pump system to adjust the air bag pressure.

7. The method for adaptive adjustment of sitting posture based on multi-modal data according to claim 1, characterized in that: In the process of determining a current posture of a user based on the pressure distribution data and the torso posture angle data, a confidence score of the current posture of the user is obtained, and when the confidence score corresponding to the posture continuously exceeds a preset threshold for a preset time, the next step of sitting posture execution action is performed.

8. The method for adaptive adjustment of sitting posture based on multi-modal data according to claim 7, characterized in that: When the difference between the confidence scores corresponding to the two postures with the highest confidence scores is lower than a preset value, it is judged that this is a posture transition state, and a preset transition execution scheme is executed.

9. A sitting posture self-adaptive adjustment system based on multi-modal data, characterized in that, The system comprises: a data acquisition module configured to collect a pressure distribution receipt through a multi-dimensional pressure array sensor, and collect torso posture angle data through an inertial measurement unit; a posture determination module configured to determine a current posture of a user based on the pressure distribution data and the torso posture angle data; a strategy execution module configured to perform a posture strategy corresponding to the current posture according to a predicted strategy rule based on the current posture of the user.

10. A seat comprising a main frame on which a support body is provided, characterized by, The support body is provided with a multi-dimensional pressure array sensor and an inertial measurement unit. The seat further comprises a memory, and the memory stores a computer program. When the computer program is executed by a processor, the method according to any one of claims 1-8 is implemented.

Images